metadata
license: mit
tags:
- generated_from_trainer
metrics:
- precision
- recall
- f1
- accuracy
model-index:
- name: model_6012023
results: []
model_6012023
This model is a fine-tuned version of Buseak/my_pos_model on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.2391
- Precision: 0.9109
- Recall: 0.9042
- F1: 0.9076
- Accuracy: 0.9348
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
- num_epochs: 3
Training results
| Training Loss | Epoch | Step | Validation Loss | Precision | Recall | F1 | Accuracy |
|---|---|---|---|---|---|---|---|
| No log | 1.0 | 244 | 0.3030 | 0.8906 | 0.8845 | 0.8875 | 0.9202 |
| No log | 2.0 | 488 | 0.2526 | 0.9051 | 0.8977 | 0.9014 | 0.9306 |
| 0.4278 | 3.0 | 732 | 0.2391 | 0.9109 | 0.9042 | 0.9076 | 0.9348 |
Framework versions
- Transformers 4.25.1
- Pytorch 1.13.0+cu116
- Datasets 2.8.0
- Tokenizers 0.13.2