|
|
--- |
|
|
license: apache-2.0 |
|
|
tags: |
|
|
- generated_from_trainer |
|
|
- biology |
|
|
datasets: |
|
|
- bionlp2004 |
|
|
model-index: |
|
|
- name: bert-base-cased-finetuned-ner-bio_nlp_2004 |
|
|
results: [] |
|
|
language: |
|
|
- en |
|
|
metrics: |
|
|
- seqeval |
|
|
- f1 |
|
|
- recall |
|
|
- precision |
|
|
pipeline_tag: token-classification |
|
|
--- |
|
|
|
|
|
# bert-base-cased-finetuned-ner-bio_nlp_2004 |
|
|
|
|
|
This model is a fine-tuned version of [bert-base-cased](https://huggingface.co/bert-base-cased). |
|
|
|
|
|
It achieves the following results on the evaluation set: |
|
|
- Loss: 0.2066 |
|
|
- Dna: |
|
|
- Precision: 0.6619127516778524 |
|
|
- Recall: 0.7471590909090909 |
|
|
- F1: 0.7019572953736656 |
|
|
- Number: 1056 |
|
|
- Rna: |
|
|
- Precision: 0.589041095890411 |
|
|
- Recall: 0.7288135593220338 |
|
|
- F1: 0.6515151515151515 |
|
|
- Number': 118 |
|
|
- Cell Line: |
|
|
- Precision: 0.4758522727272727 |
|
|
- Recall: 0.67 |
|
|
- F1: 0.5564784053156145 |
|
|
- Number: 500 |
|
|
- Cell Type: |
|
|
- Precision: 0.7294117647058823 |
|
|
- Recall: 0.7100468505986466 |
|
|
- F1: 0.7195990503824848 |
|
|
- Number: 1921 |
|
|
- Protein: |
|
|
- Precision: 0.6657656225155033 |
|
|
- Recall: 0.8263272153147819 |
|
|
- F1: 0.7374075378654457 |
|
|
- Number': 5067 |
|
|
|
|
|
- Overall |
|
|
- Precision: 0.6628 |
|
|
- Recall: 0.7805 |
|
|
- F1: 0.7169 |
|
|
- Accuracy: 0.9367 |
|
|
|
|
|
## Model description |
|
|
|
|
|
For more information on how it was created, check out the following link: https://github.com/DunnBC22/NLP_Projects/blob/main/Token%20Classification/Monolingual/tner-bionlp2004/NER%20Project%20Using%20tner-bionlp%202004%20Dataset%20(BERT-Base).ipynb |
|
|
|
|
|
## Intended uses & limitations |
|
|
|
|
|
This model is intended to demonstrate my ability to solve a complex problem using technology. |
|
|
|
|
|
## Training and evaluation data |
|
|
|
|
|
Dataset Source: https://huggingface.co/datasets/tner/bionlp2004 |
|
|
|
|
|
## 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 |
|
|
|
|
|
### Training results |
|
|
| Training Loss | Epoch | Step | Valid. Loss | Dna Precision | Dna Recall | Dna F1 | Dna Number | Rna Precision | Rna Recall | Rna F1 | Rna Number | Cell Line Precision | Cell Type Recall | Cell Type F1 | Cell Type Number | Cell Type | Protein Precision | Protein Recall | Protein F1 | Protein Number | Overall Precision | Overall Recall | Overall F1 | Overall Accuracy | |
|
|
|:---------:|:-----:|:----:|:---------:|:-------:|:------:|:------:|:------:|:-------------:|:----------:|:------:|:----------:|:----------:|:---------:|:-------:|:-------:|:---------:|:-------:|:--------:|:------:|:-----------:|:--------:|:--------:|:----------:|:---------:| |
|
|
| 0.1701 | 1.0 | 1039 | 0.1927 | 0.6153 | 0.7254 | 0.6658 | 1056 | 0.6617 | 0.7458 | 0.7012 | 118 | 0.4670 | 0.608 | 0.5282 | 500 | 0.6997 | 0.7158 | 0.7077 | 1921 | 0.6603 | 0.7833 | 0.7166 | 5067 | 0.6499 | 0.7506 | 0.6966 | 0.9352 | |
|
|
| 0.145 | 2.0 | 2078 | 0.1981 | 0.6364 | 0.7443 | 0.6862 | 1056 | 0.6408 | 0.7712 | 0.7000 | 118 | 0.4607 | 0.668 | 0.5453 | 500 | 0.7376 | 0.7022 | 0.7195 | 1921 | 0.6759 | 0.8149 | 0.7389 | 5067 | 0.6662 | 0.7722 | 0.7153 | 0.9364 | |
|
|
| 0.1116 | 3.0 | 3117 | 0.2066 | 0.6619 | 0.7472 | 0.7020 | 1056 | 0.5890 | 0.7288 | 0.6515 | 118 | 0.4759 | 0.67 | 0.5565 | 500 | 0.7294 | 0.7100 | 0.7196 | 1921 | 0.6658 | 0.8263 | 0.7374 | 5067 | 0.6628 | 0.7805 | 0.7169 | 0.9367 | |
|
|
|
|
|
* Metrics shown above are rounded to the neareset ten-thousandth |
|
|
|
|
|
### Framework versions |
|
|
|
|
|
- Transformers 4.28.1 |
|
|
- Pytorch 2.0.0 |
|
|
- Datasets 2.11.0 |
|
|
- Tokenizers 0.13.3 |
