| --- |
| license: mit |
| base_model: gpt2 |
| tags: |
| - generated_from_trainer |
| datasets: |
| - wikitext |
| model-index: |
| - name: gpt2-wikitext2 |
| results: [] |
| --- |
| |
| <!-- This model card has been generated automatically according to the information the Trainer had access to. You |
| should probably proofread and complete it, then remove this comment. --> |
|
|
| # gpt2-wikitext2 |
|
|
| This model is a fine-tuned version of [gpt2](https://huggingface.co/gpt2) on the wikitext dataset. |
| It achieves the following results on the evaluation set: |
| - Loss: 6.1045 |
|
|
| ## 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: 8 |
| - eval_batch_size: 8 |
| - seed: 42 |
| - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 |
| - lr_scheduler_type: linear |
| - num_epochs: 3.0 |
|
|
| ### Training results |
|
|
| | Training Loss | Epoch | Step | Validation Loss | |
| |:-------------:|:-----:|:----:|:---------------:| |
| | 6.5489 | 1.0 | 2250 | 6.4641 | |
| | 6.148 | 2.0 | 4500 | 6.1911 | |
| | 6.0043 | 3.0 | 6750 | 6.1045 | |
|
|
|
|
| ### Framework versions |
|
|
| - Transformers 4.32.0.dev0 |
| - Pytorch 2.0.1+cpu |
| - Datasets 2.12.0 |
| - Tokenizers 0.13.2 |
|
|
