|
--- |
|
license: apache-2.0 |
|
tags: |
|
- generated_from_trainer |
|
metrics: |
|
- precision |
|
- accuracy |
|
- f1 |
|
model-index: |
|
- name: TESTING |
|
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. --> |
|
|
|
# TESTING |
|
|
|
This model is a fine-tuned version of [distilroberta-base](https://huggingface.co/distilroberta-base) on an unknown dataset. |
|
It achieves the following results on the evaluation set: |
|
- Loss: 0.1167 |
|
- Precision: 0.9561 |
|
- Accuracy: 0.9592 |
|
- F1: 0.9592 |
|
|
|
## 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: 32 |
|
- eval_batch_size: 32 |
|
- seed: 42 |
|
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 |
|
- lr_scheduler_type: linear |
|
- lr_scheduler_warmup_steps: 500 |
|
- num_epochs: 2 |
|
|
|
### Training results |
|
|
|
| Training Loss | Epoch | Step | Validation Loss | Precision | Accuracy | F1 | |
|
|:-------------:|:-----:|:----:|:---------------:|:---------:|:--------:|:------:| |
|
| 0.5903 | 0.4 | 500 | 0.4695 | 0.7342 | 0.7728 | 0.7890 | |
|
| 0.3986 | 0.8 | 1000 | 0.3469 | 0.8144 | 0.8596 | 0.8684 | |
|
| 0.2366 | 1.2 | 1500 | 0.1939 | 0.9313 | 0.9260 | 0.9253 | |
|
| 0.1476 | 1.6 | 2000 | 0.1560 | 0.9207 | 0.9452 | 0.9465 | |
|
| 0.1284 | 2.0 | 2500 | 0.1167 | 0.9561 | 0.9592 | 0.9592 | |
|
|
|
|
|
### Framework versions |
|
|
|
- Transformers 4.18.0 |
|
- Pytorch 1.10.0+cu111 |
|
- Datasets 2.1.0 |
|
- Tokenizers 0.12.1 |
|
|
