Anti-overfitting 5-class sentiment model

README.md

@@ -18,10 +18,10 @@ should probably proofread and complete it, then remove this comment. -->
 
 This model is a fine-tuned version of [cardiffnlp/twitter-roberta-base-sentiment-latest](https://huggingface.co/cardiffnlp/twitter-roberta-base-sentiment-latest) on the None dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.1496
-- Accuracy: 0.956
-- F1 Macro: 0.9544
-- F1 Weighted: 0.9553
+- Loss: 0.1090
+- Accuracy: 0.9668
+- F1 Macro: 0.9673
+- F1 Weighted: 0.9669
 
 ## Model description
 
@@ -54,14 +54,14 @@ The following hyperparameters were used during training:
 
 ### Training results
 
-| Training Loss | Epoch | Step | Validation Loss | Accuracy | F1 Macro | F1 Weighted |
-|:-------------:|:-----:|:----:|:---------------:|:--------:|:--------:|:-----------:|
-| 1.3595        | 0.16  | 20   | 1.1077          | 0.594    | 0.5350   | 0.5278      |
-| 0.9369        | 0.32  | 40   | 0.5998          | 0.822    | 0.8158   | 0.8165      |
-| 0.6807        | 0.48  | 60   | 0.3187          | 0.93     | 0.9277   | 0.9292      |
-| 0.4498        | 0.64  | 80   | 0.2548          | 0.92     | 0.9189   | 0.9192      |
-| 0.3857        | 0.8   | 100  | 0.1801          | 0.95     | 0.9490   | 0.9499      |
-| 0.3638        | 0.96  | 120  | 0.1496          | 0.956    | 0.9544   | 0.9553      |
+| Training Loss | Epoch  | Step | Validation Loss | Accuracy | F1 Macro | F1 Weighted |
+|:-------------:|:------:|:----:|:---------------:|:--------:|:--------:|:-----------:|
+| 1.2893        | 0.1559 | 20   | 0.9350          | 0.6074   | 0.4810   | 0.4802      |
+| 0.8368        | 0.3119 | 40   | 0.5051          | 0.8848   | 0.8833   | 0.8831      |
+| 0.6255        | 0.4678 | 60   | 0.2471          | 0.9336   | 0.9341   | 0.9336      |
+| 0.469         | 0.6238 | 80   | 0.1967          | 0.9297   | 0.9299   | 0.9295      |
+| 0.3423        | 0.7797 | 100  | 0.1227          | 0.9551   | 0.9558   | 0.9553      |
+| 0.3477        | 0.9357 | 120  | 0.1090          | 0.9668   | 0.9673   | 0.9669      |
 
 
 ### Framework versions