metadata
tags:
- generated_from_trainer
metrics:
- accuracy
model-index:
- name: BERTClassificationBaseline
results: []
BERTClassificationBaseline
This model is a fine-tuned version of allenai/scibert_scivocab_uncased on the None dataset. It achieves the following results on the evaluation set:
- Loss: 1.2709
- Accuracy: 0.4
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-06
- train_batch_size: 4
- eval_batch_size: 4
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 2
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy |
|---|---|---|---|---|
| No log | 1.0 | 157 | 1.3100 | 0.4286 |
| No log | 2.0 | 314 | 1.2709 | 0.4 |
Framework versions
- Transformers 4.22.1
- Pytorch 1.12.1
- Datasets 2.5.2
- Tokenizers 0.11.0