bert_uncased_L-4_H-128_A-2-mlm-multi-emails-hq
This model is a fine-tuned version of google/bert_uncased_L-4_H-128_A-2 on the None dataset. It achieves the following results on the evaluation set:
- Loss: 2.8524
- Accuracy: 0.5077
Model description
Double the layers of BERT-tiny, fine-tuned on email data for eight epochs.
Intended uses & limitations
- This is primarily an example/test
Training and evaluation data
More information needed
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 0.0003
- train_batch_size: 8
- eval_batch_size: 2
- seed: 42
- distributed_type: multi-GPU
- gradient_accumulation_steps: 16
- total_train_batch_size: 128
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_ratio: 0.05
- num_epochs: 8.0
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy |
|---|---|---|---|---|
| 3.5477 | 0.99 | 141 | 3.2637 | 0.4551 |
| 3.3307 | 1.99 | 282 | 3.0873 | 0.4785 |
| 3.252 | 2.99 | 423 | 2.9842 | 0.4911 |
| 3.1415 | 3.99 | 564 | 2.9230 | 0.4995 |
| 3.0903 | 4.99 | 705 | 2.8625 | 0.5070 |
| 3.0996 | 5.99 | 846 | 2.8615 | 0.5087 |
| 3.0641 | 6.99 | 987 | 2.8407 | 0.5120 |
| 3.0514 | 7.99 | 1128 | 2.8524 | 0.5077 |
Framework versions
- Transformers 4.27.0.dev0
- Pytorch 2.0.0.dev20230129+cu118
- Datasets 2.8.0
- Tokenizers 0.13.1