metadata
license: apache-2.0
tags:
- generated_from_trainer
datasets:
- data_set
metrics:
- precision
- recall
- f1
- accuracy
model-index:
- name: finetuned-ner-offer
results:
- task:
name: Token Classification
type: token-classification
dataset:
name: data_set
type: data_set
config: conll2003
split: train
args: conll2003
metrics:
- name: Precision
type: precision
value: 0.2254335260115607
- name: Recall
type: recall
value: 0.24528301886792453
- name: F1
type: f1
value: 0.23493975903614459
- name: Accuracy
type: accuracy
value: 0.9209345919528688
finetuned-ner-offer
This model is a fine-tuned version of bert-base-cased on the data_set dataset. It achieves the following results on the evaluation set:
- Loss: 0.3242
- Precision: 0.2254
- Recall: 0.2453
- F1: 0.2349
- Accuracy: 0.9209
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 |
|---|---|---|---|---|---|---|---|
| No log | 1.0 | 100 | 0.3344 | 0.2162 | 0.2013 | 0.2085 | 0.9217 |
| No log | 2.0 | 200 | 0.3167 | 0.1804 | 0.2201 | 0.1983 | 0.9204 |
| No log | 3.0 | 300 | 0.3242 | 0.2254 | 0.2453 | 0.2349 | 0.9209 |
Framework versions
- Transformers 4.25.1
- Pytorch 1.13.0+cu116
- Datasets 2.8.0
- Tokenizers 0.13.2