Llama3_2_3b_stage1
This model is a fine-tuned version of meta-llama/Llama-3.2-3B-Instruct on the None dataset. It achieves the following results on the evaluation set:
- Loss: 4.7297
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: 1
- eval_batch_size: 1
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 4
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 40
- num_epochs: 2
Training results
| Training Loss | Epoch | Step | Validation Loss |
|---|---|---|---|
| 10.1992 | 0.13 | 1000 | 10.5695 |
| 10.7364 | 0.26 | 2000 | 11.1818 |
| 9.3268 | 0.39 | 3000 | 10.7739 |
| 10.3635 | 0.52 | 4000 | 10.3186 |
| 7.9656 | 0.65 | 5000 | 9.5328 |
| 7.7822 | 0.78 | 6000 | 8.7203 |
| 8.8225 | 0.91 | 7000 | 8.1419 |
| 5.6131 | 1.04 | 8000 | 7.3168 |
| 5.7632 | 1.16 | 9000 | 6.5680 |
| 5.7977 | 1.29 | 10000 | 5.9854 |
| 5.3294 | 1.42 | 11000 | 5.4301 |
| 4.4568 | 1.55 | 12000 | 5.0660 |
| 4.7171 | 1.68 | 13000 | 4.8114 |
| 4.6005 | 1.81 | 14000 | 4.7361 |
| 4.4726 | 1.94 | 15000 | 4.7297 |
Framework versions
- Transformers 4.39.1
- Pytorch 2.6.0+cu126
- Datasets 2.18.0
- Tokenizers 0.15.2
- Downloads last month
- 4
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support
Model tree for Michael-Shin/Medusa_test_5_heads
Base model
meta-llama/Llama-3.2-3B-Instruct