---
license: mit
datasets:
- yahma/alpaca-cleaned
---

This repo reproduced [tloen/alpaca-lora-7b](https://huggingface.co/tloen/alpaca-lora-7b)
fit on the [Stanford Alpaca](https://github.com/tatsu-lab/stanford_alpaca) dataset.

4x H100 training for about 1h15min, details in [W&B link](https://wandb.ai/sharpbai/alpaca-lora-reproduce/runs/08ulvstd), there is a hyperparameter of val_set_size=500

4 x 4090 training for about 4h35min, details in [W&B link](https://wandb.ai/sharpbai/alpaca-lora-reproduce/runs/ws16av1u), all key hyperparameters are the same

To optimize the running speed, I change these code

- `load_in_8bits=False` to use 16bit finetune
- comment `model = prepare_model_for_int8_training` to not turn some parameters to fp32 and turn off gradient checkpointing
- for 4090 enable gradient checkpointing, add `model.gradient_checkpointing_enable()` and `model.enable_input_require_grads()`

This version of the weights was trained with the following hyperparameters:

- Epochs: 10 (load from best epoch)
- Batch size: 128
- Cutoff length: 512
- Learning rate: 3e-4
- Lora _r_: 16
- Lora target modules: q_proj, k_proj, v_proj, o_proj

That is:

```
python finetune.py \
    --base_model='decapoda-research/llama-7b-hf' \
    --num_epochs=10 \
    --cutoff_len=512 \
    --group_by_length \
    --val_set_size=500 \
    --output_dir='./alpaca-lora-train-H100-80G-HBM3x4-mb8' \
    --lora_target_modules='[q_proj,k_proj,v_proj,o_proj]' \
    --lora_r=16 \
    --micro_batch_size=8 \
    --train_in_8bit False
```

Instructions for running it can be found at https://github.com/tloen/alpaca-lora.