|
--- |
|
license: apache-2.0 |
|
tags: |
|
- generated_from_trainer |
|
metrics: |
|
- accuracy |
|
- f1 |
|
model-index: |
|
- name: bert-nlp-project-ft-google-ds-imdb |
|
results: [] |
|
--- |
|
|
|
<!-- This model card has been generated automatically according to the information the Trainer had access to. You |
|
should probably proofread and complete it, then remove this comment. --> |
|
|
|
# bert-nlp-project-ft-google-ds-imdb |
|
|
|
This model is a fine-tuned version of [jestemleon/bert-nlp-project-google](https://huggingface.co/jestemleon/bert-nlp-project-google) on the None dataset. |
|
It achieves the following results on the evaluation set: |
|
- Loss: 0.2646 |
|
- Accuracy: 0.9455 |
|
- F1: 0.9446 |
|
|
|
## Model description |
|
|
|
More information needed |
|
|
|
## Intended uses & limitations |
|
|
|
More information needed |
|
|
|
## Training and evaluation data |
|
|
|
More information needed |
|
|
|
## Training procedure |
|
|
|
### Training hyperparameters |
|
|
|
The following hyperparameters were used during training: |
|
- learning_rate: 2e-05 |
|
- train_batch_size: 16 |
|
- eval_batch_size: 16 |
|
- seed: 42 |
|
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 |
|
- lr_scheduler_type: linear |
|
- num_epochs: 3 |
|
- mixed_precision_training: Native AMP |
|
|
|
### Training results |
|
|
|
| Training Loss | Epoch | Step | Validation Loss | Accuracy | F1 | |
|
|:-------------:|:-----:|:----:|:---------------:|:--------:|:------:| |
|
| 0.2684 | 0.38 | 750 | 0.1850 | 0.9315 | 0.9295 | |
|
| 0.2105 | 0.75 | 1500 | 0.1734 | 0.9407 | 0.9385 | |
|
| 0.1767 | 1.12 | 2250 | 0.2229 | 0.9337 | 0.9346 | |
|
| 0.1259 | 1.5 | 3000 | 0.2029 | 0.9423 | 0.9401 | |
|
| 0.1187 | 1.88 | 3750 | 0.2554 | 0.9353 | 0.9359 | |
|
| 0.0778 | 2.25 | 4500 | 0.2759 | 0.9433 | 0.9418 | |
|
| 0.06 | 2.62 | 5250 | 0.2663 | 0.9443 | 0.9428 | |
|
| 0.0551 | 3.0 | 6000 | 0.2646 | 0.9455 | 0.9446 | |
|
|
|
|
|
### Framework versions |
|
|
|
- Transformers 4.25.1 |
|
- Pytorch 1.13.0+cu116 |
|
- Datasets 2.8.0 |
|
- Tokenizers 0.13.2 |
|
|