|
--- |
|
tags: |
|
- generated_from_trainer |
|
datasets: |
|
- HiTZ/alpaca_mt |
|
model-index: |
|
- name: alpaca-lora-65b-en-pt-es-ca |
|
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. --> |
|
|
|
# alpaca-lora-65b-en-pt-es-ca |
|
|
|
This model is a fine-tuned version of [/gaueko1/hizkuntza-ereduak/LLaMA/lm/huggingface/65B](https://huggingface.co//gaueko1/hizkuntza-ereduak/LLaMA/lm/huggingface/65B) on the HiTZ/alpaca_mt ['en', 'pt', 'es', 'ca'] dataset. |
|
It achieves the following results on the evaluation set: |
|
- Loss: 0.7271 |
|
|
|
## 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 |
|
- num_devices: 2 |
|
- gradient_accumulation_steps: 63 |
|
- total_train_batch_size: 126 |
|
- total_eval_batch_size: 2 |
|
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 |
|
- lr_scheduler_type: cosine |
|
- lr_scheduler_warmup_ratio: 0.03 |
|
- num_epochs: 1 |
|
- mixed_precision_training: Native AMP |
|
|
|
### Training results |
|
|
|
| Training Loss | Epoch | Step | Validation Loss | |
|
|:-------------:|:-----:|:----:|:---------------:| |
|
| 0.8069 | 0.06 | 100 | 0.8033 | |
|
| 0.8008 | 0.13 | 200 | 0.7826 | |
|
| 0.7687 | 0.19 | 300 | 0.7721 | |
|
| 0.7719 | 0.25 | 400 | 0.7647 | |
|
| 0.7585 | 0.32 | 500 | 0.7588 | |
|
| 0.7578 | 0.38 | 600 | 0.7537 | |
|
| 0.7505 | 0.44 | 700 | 0.7491 | |
|
| 0.7531 | 0.51 | 800 | 0.7449 | |
|
| 0.7394 | 0.57 | 900 | 0.7416 | |
|
| 0.7368 | 0.63 | 1000 | 0.7387 | |
|
| 0.7412 | 0.69 | 1100 | 0.7361 | |
|
| 0.7344 | 0.76 | 1200 | 0.7288 | |
|
| 0.7383 | 0.82 | 1300 | 0.7281 | |
|
| 0.7378 | 0.88 | 1400 | 0.7274 | |
|
| 0.7204 | 0.95 | 1500 | 0.7271 | |
|
|
|
|
|
### Framework versions |
|
|
|
- Transformers 4.28.0.dev0 |
|
- Pytorch 2.0.0+cu117 |
|
- Datasets 2.10.1 |
|
- Tokenizers 0.13.2 |
|
|
|
# [Open LLM Leaderboard Evaluation Results](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard) |
|
Detailed results can be found [here](https://huggingface.co/datasets/open-llm-leaderboard/details_HiTZ__alpaca-lora-65b-en-pt-es-ca) |
|
|
|
| Metric | Value | |
|
|-----------------------|---------------------------| |
|
| Avg. | 59.31 | |
|
| ARC (25-shot) | 65.02 | |
|
| HellaSwag (10-shot) | 84.88 | |
|
| MMLU (5-shot) | 62.19 | |
|
| TruthfulQA (0-shot) | 46.06 | |
|
| Winogrande (5-shot) | 80.51 | |
|
| GSM8K (5-shot) | 26.69 | |
|
| DROP (3-shot) | 49.84 | |
|
|