metadata
license: llama2
base_model: meta-llama/Llama-2-7b-hf
tags:
- generated_from_trainer
datasets:
- mozilla-foundation/common_voice_11_0
model-index:
- name: laft_lug_llama
results: []
laft_lug_llama
This model is a fine-tuned version of meta-llama/Llama-2-7b-hf on the mozilla-foundation/common_voice_11_0 lg dataset. It achieves the following results on the evaluation set:
- Loss: 3.3325
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: 0.0003
- train_batch_size: 1
- eval_batch_size: 1
- seed: 42
- distributed_type: multi-GPU
- optimizer: Adam with betas=(0.9,0.95) and epsilon=1e-05
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 2000
- num_epochs: 6.0
Training results
| Training Loss | Epoch | Step | Validation Loss |
|---|---|---|---|
| 2.7093 | 1.0 | 1599 | 2.3153 |
| 2.6641 | 2.0 | 3198 | 2.1519 |
| 1.8919 | 3.0 | 4797 | 1.9803 |
| 1.5401 | 4.0 | 6396 | 2.0956 |
| 0.5044 | 5.0 | 7995 | 2.7269 |
| 0.1869 | 6.0 | 9594 | 3.3325 |
Framework versions
- Transformers 4.44.0
- Pytorch 2.4.0+cu121
- Datasets 2.20.0
- Tokenizers 0.19.1