metadata
license: llama2
base_model: lmsys/vicuna-7b-v1.5
tags:
- generated_from_trainer
datasets:
- truthful_qa
metrics:
- accuracy
model-index:
- name: vicuna_mc_finetune
results: []
vicuna_mc_finetune
This model is a fine-tuned version of lmsys/vicuna-7b-v1.5 on the truthful_qa dataset. It achieves the following results on the evaluation set:
- Loss: 1.8867
- Accuracy: 0.2378
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: 0.0001
- train_batch_size: 6
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 5
- num_epochs: 20
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy |
|---|---|---|---|---|
| 1.7307 | 1.0 | 109 | 1.5327 | 0.4024 |
| 1.4519 | 2.0 | 218 | 1.2341 | 0.7073 |
| 0.0207 | 3.0 | 327 | 2.2924 | 0.6280 |
| 1.5647 | 4.0 | 436 | 1.9344 | 0.2256 |
| 2.2047 | 5.0 | 545 | 1.9401 | 0.2256 |
| 2.016 | 6.0 | 654 | 1.8888 | 0.2256 |
| 1.625 | 7.0 | 763 | 1.9068 | 0.1768 |
| 2.0002 | 8.0 | 872 | 1.8909 | 0.1951 |
| 1.7906 | 9.0 | 981 | 1.8828 | 0.2195 |
| 1.5295 | 10.0 | 1090 | 1.8967 | 0.2195 |
| 1.7018 | 11.0 | 1199 | 1.8845 | 0.2378 |
| 1.8412 | 12.0 | 1308 | 1.8808 | 0.2073 |
| 2.4396 | 13.0 | 1417 | 1.8816 | 0.2012 |
| 1.8643 | 14.0 | 1526 | 1.8827 | 0.2012 |
| 1.7271 | 15.0 | 1635 | 1.8844 | 0.2256 |
| 1.851 | 16.0 | 1744 | 1.8720 | 0.2134 |
| 1.7633 | 17.0 | 1853 | 1.8786 | 0.2134 |
| 2.6586 | 18.0 | 1962 | 1.8723 | 0.25 |
| 2.0078 | 19.0 | 2071 | 1.8770 | 0.2439 |
| 1.4072 | 20.0 | 2180 | 1.8867 | 0.2378 |
Framework versions
- Transformers 4.36.0.dev0
- Pytorch 2.1.0+cu121
- Datasets 2.13.1
- Tokenizers 0.14.1