metadata
license: mit
tags:
- generated_from_trainer
metrics:
- precision
- recall
- f1
- accuracy
model-index:
- name: model_from_berturk_1401_v2
results: []
model_from_berturk_1401_v2
This model is a fine-tuned version of Buseak/model_from_berturk_1401 on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.1542
- Precision: 0.9414
- Recall: 0.9356
- F1: 0.9385
- Accuracy: 0.9569
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: 4
Training results
| Training Loss | Epoch | Step | Validation Loss | Precision | Recall | F1 | Accuracy |
|---|---|---|---|---|---|---|---|
| No log | 1.0 | 244 | 0.2277 | 0.9129 | 0.9058 | 0.9094 | 0.9362 |
| No log | 2.0 | 488 | 0.1855 | 0.9275 | 0.9204 | 0.9240 | 0.9472 |
| 0.2477 | 3.0 | 732 | 0.1602 | 0.9403 | 0.9315 | 0.9359 | 0.9554 |
| 0.2477 | 4.0 | 976 | 0.1542 | 0.9414 | 0.9356 | 0.9385 | 0.9569 |
Framework versions
- Transformers 4.25.1
- Pytorch 1.13.0+cu116
- Datasets 2.8.0
- Tokenizers 0.13.2