LLama_End
This model is a fine-tuned version of /teamspace/studios/this_studio/save/SFT_merge on the WordProblems_SFT_LLama_End dataset. It achieves the following results on the evaluation set:
- Loss: 0.0769
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: 5e-05
- train_batch_size: 4
- eval_batch_size: 4
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 16
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 500
- num_epochs: 2.0
Training results
| Training Loss | Epoch | Step | Validation Loss |
|---|---|---|---|
| 0.1013 | 0.2722 | 500 | 0.0977 |
| 0.0851 | 0.5444 | 1000 | 0.0865 |
| 0.0819 | 0.8167 | 1500 | 0.0826 |
| 0.0753 | 1.0889 | 2000 | 0.0804 |
| 0.0737 | 1.3611 | 2500 | 0.0786 |
| 0.0692 | 1.6333 | 3000 | 0.0773 |
| 0.0671 | 1.9055 | 3500 | 0.0769 |
Framework versions
- PEFT 0.12.0
- Transformers 4.43.4
- Pytorch 2.2.1+cu121
- Datasets 2.20.0
- Tokenizers 0.19.1