Mlukebasewithpaperparameter_learningrate2e-5
This model is a fine-tuned version of studio-ousia/mluke-base on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.1361
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: 12
- eval_batch_size: 24
- seed: 42
- gradient_accumulation_steps: 5
- total_train_batch_size: 60
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.005
- num_epochs: 10
Training results
Training Loss | Epoch | Step | Validation Loss |
---|---|---|---|
0.3659 | 0.9928 | 55 | 0.3273 |
0.2429 | 1.9856 | 110 | 0.2136 |
0.2394 | 2.9964 | 166 | 0.1950 |
0.1483 | 3.9892 | 221 | 0.1818 |
0.1393 | 5.0 | 277 | 0.1383 |
0.126 | 5.9928 | 332 | 0.1482 |
0.1 | 6.9856 | 387 | 0.1480 |
0.1042 | 7.9964 | 443 | 0.1297 |
0.078 | 8.9892 | 498 | 0.1270 |
0.0961 | 9.9278 | 550 | 0.1361 |
Framework versions
- Transformers 4.40.1
- Pytorch 2.2.1+cu121
- Datasets 2.19.1
- Tokenizers 0.19.1
- Downloads last month
- 4