ai-research-lab/bert-question-classifier

README.md

@@ -21,11 +21,11 @@ should probably proofread and complete it, then remove this comment. -->
 
 This model is a fine-tuned version of [google-bert/bert-base-uncased](https://huggingface.co/google-bert/bert-base-uncased) on the None dataset.
 It achieves the following results on the evaluation set:
-- Loss: 2.7007
-- Accuracy: 0.9562
-- Recall: 0.7745
-- Precision: 0.7606
-- F1: 0.7675
+- Loss: 1.8460
+- Accuracy: 0.9711
+- Recall: 0.8571
+- Precision: 0.8371
+- F1: 0.8470
 
 ## Model description
 
@@ -44,19 +44,42 @@ More information needed
 ### Training hyperparameters
 
 The following hyperparameters were used during training:
-- learning_rate: 2e-05
-- train_batch_size: 8
-- eval_batch_size: 8
+- learning_rate: 1e-05
+- train_batch_size: 4
+- eval_batch_size: 4
 - seed: 42
 - optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
 - lr_scheduler_type: linear
-- num_epochs: 1
+- lr_scheduler_warmup_ratio: 0.1
+- num_epochs: 3
 
 ### Training results
 
-| Training Loss | Epoch | Step | Validation Loss | Accuracy | Recall | Precision | F1     |
-|:-------------:|:-----:|:----:|:---------------:|:--------:|:------:|:---------:|:------:|
-| No log        | 1.0   | 390  | 2.7007          | 0.9562   | 0.7745 | 0.7606    | 0.7675 |
+| Training Loss | Epoch  | Step | Validation Loss | Accuracy | Recall | Precision | F1     |
+|:-------------:|:------:|:----:|:---------------:|:--------:|:------:|:---------:|:------:|
+| No log        | 0.1284 | 100  | 4.9916          | 0.9143   | 0.5142 | 0.5430    | 0.5282 |
+| No log        | 0.2567 | 200  | 4.3263          | 0.9264   | 0.5870 | 0.6092    | 0.5979 |
+| No log        | 0.3851 | 300  | 3.8919          | 0.9348   | 0.6636 | 0.6463    | 0.6548 |
+| No log        | 0.5135 | 400  | 3.5265          | 0.9391   | 0.6599 | 0.6783    | 0.6690 |
+| 4.2904        | 0.6418 | 500  | 3.2937          | 0.9452   | 0.7049 | 0.7066    | 0.7057 |
+| 4.2904        | 0.7702 | 600  | 3.0129          | 0.9496   | 0.7275 | 0.7305    | 0.7290 |
+| 4.2904        | 0.8986 | 700  | 2.8410          | 0.9521   | 0.7482 | 0.7404    | 0.7443 |
+| 4.2904        | 1.0270 | 800  | 2.6565          | 0.9552   | 0.7757 | 0.7520    | 0.7637 |
+| 4.2904        | 1.1553 | 900  | 2.5233          | 0.9574   | 0.7842 | 0.7647    | 0.7743 |
+| 2.7537        | 1.2837 | 1000 | 2.3877          | 0.9598   | 0.7976 | 0.7771    | 0.7872 |
+| 2.7537        | 1.4121 | 1100 | 2.2836          | 0.9622   | 0.8146 | 0.7875    | 0.8008 |
+| 2.7537        | 1.5404 | 1200 | 2.1776          | 0.9635   | 0.8130 | 0.7990    | 0.8059 |
+| 2.7537        | 1.6688 | 1300 | 2.1273          | 0.9653   | 0.8223 | 0.8085    | 0.8153 |
+| 2.7537        | 1.7972 | 1400 | 2.0858          | 0.9651   | 0.8251 | 0.8052    | 0.8150 |
+| 2.218         | 1.9255 | 1500 | 2.0143          | 0.9670   | 0.8312 | 0.8176    | 0.8243 |
+| 2.218         | 2.0539 | 1600 | 1.9800          | 0.9683   | 0.8413 | 0.8226    | 0.8319 |
+| 2.218         | 2.1823 | 1700 | 1.9409          | 0.9691   | 0.8470 | 0.8259    | 0.8363 |
+| 2.218         | 2.3107 | 1800 | 1.9122          | 0.9693   | 0.8445 | 0.8294    | 0.8369 |
+| 2.218         | 2.4390 | 1900 | 1.8876          | 0.9699   | 0.8502 | 0.8310    | 0.8405 |
+| 1.8644        | 2.5674 | 2000 | 1.8786          | 0.9695   | 0.8470 | 0.8298    | 0.8383 |
+| 1.8644        | 2.6958 | 2100 | 1.8620          | 0.9701   | 0.8510 | 0.8325    | 0.8416 |
+| 1.8644        | 2.8241 | 2200 | 1.8494          | 0.9711   | 0.8567 | 0.8370    | 0.8467 |
+| 1.8644        | 2.9525 | 2300 | 1.8460          | 0.9711   | 0.8571 | 0.8371    | 0.8470 |
 
 
 ### Framework versions