|
--- |
|
license: apache-2.0 |
|
tags: |
|
- generated_from_trainer |
|
datasets: |
|
- conll2003 |
|
metrics: |
|
- precision |
|
- recall |
|
- f1 |
|
- accuracy |
|
model-index: |
|
- name: bert-finetuned-ner3 |
|
results: |
|
- task: |
|
name: Token Classification |
|
type: token-classification |
|
dataset: |
|
name: conll2003 |
|
type: conll2003 |
|
args: conll2003 |
|
metrics: |
|
- name: Precision |
|
type: precision |
|
value: 0.9296076491922189 |
|
- name: Recall |
|
type: recall |
|
value: 0.9490070683271625 |
|
- name: F1 |
|
type: f1 |
|
value: 0.9392071952031978 |
|
- name: Accuracy |
|
type: accuracy |
|
value: 0.986342497203744 |
|
--- |
|
|
|
<!-- 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-finetuned-ner3 |
|
|
|
This model is a fine-tuned version of [bert-base-cased](https://huggingface.co/bert-base-cased) on the conll2003 dataset. |
|
It achieves the following results on the evaluation set: |
|
- Loss: 0.0603 |
|
- Precision: 0.9296 |
|
- Recall: 0.9490 |
|
- F1: 0.9392 |
|
- Accuracy: 0.9863 |
|
|
|
## 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: 8 |
|
- eval_batch_size: 8 |
|
- seed: 42 |
|
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 |
|
- lr_scheduler_type: linear |
|
- num_epochs: 3 |
|
|
|
### Training results |
|
|
|
| Training Loss | Epoch | Step | Validation Loss | Precision | Recall | F1 | Accuracy | |
|
|:-------------:|:-----:|:----:|:---------------:|:---------:|:------:|:------:|:--------:| |
|
| 0.0855 | 1.0 | 1756 | 0.0673 | 0.9130 | 0.9340 | 0.9234 | 0.9827 | |
|
| 0.0345 | 2.0 | 3512 | 0.0590 | 0.9284 | 0.9445 | 0.9363 | 0.9855 | |
|
| 0.0229 | 3.0 | 5268 | 0.0603 | 0.9296 | 0.9490 | 0.9392 | 0.9863 | |
|
|
|
|
|
### Framework versions |
|
|
|
- Transformers 4.18.0 |
|
- Pytorch 1.10.0+cu111 |
|
- Datasets 2.1.0 |
|
- Tokenizers 0.12.1 |
|
|
