| | --- |
| | license: apache-2.0 |
| | base_model: AIGym/TinyGPT2-81M |
| | tags: |
| | - generated_from_trainer |
| | model-index: |
| | - name: output |
| | 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. --> |
| |
|
| | # output |
| |
|
| | This model is a fine-tuned version of [AIGym/TinyGPT2-81M](https://huggingface.co/AIGym/TinyGPT2-81M) on an unknown dataset. |
| |
|
| | ## 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: 0.0005 |
| | - train_batch_size: 32 |
| | - eval_batch_size: 32 |
| | - seed: 42 |
| | - gradient_accumulation_steps: 8 |
| | - total_train_batch_size: 256 |
| | - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 |
| | - lr_scheduler_type: cosine |
| | - lr_scheduler_warmup_steps: 1000 |
| | - num_epochs: 1 |
| | - mixed_precision_training: Native AMP |
| |
|
| | ### Training results |
| |
|
| |
|
| |
|
| | ### Framework versions |
| |
|
| | - Transformers 4.42.4 |
| | - Pytorch 2.2.2 |
| | - Datasets 2.20.0 |
| | - Tokenizers 0.19.1 |
| |
|
