Training of RoLLMs

Official code used for training Romanian LLMs as proposed in Masala et al. 2024. This repo is a fork of the popular hiyouga/LLaMA-Factory repo used for training and finetuning LLMs. On top of the existing framework we added a suite of Romanian datasets:

• ro_sft_alpaca
• ro_sft_alpaca_gpt4
• ro_sft_dolly
• ro_sft_selfinstruct_gpt4
• ro_sft_norobts
• ro_sft_orca
• ro_sft_camel

• Requirement
• Getting Started

Requirement

Mandatory	Minimum	Recommend
python	3.8	3.11
torch	1.13.1	2.4.0
transformers	4.41.2	4.43.4
datasets	2.16.0	2.20.0
accelerate	0.30.1	0.32.0
peft	0.11.1	0.12.0
trl	0.8.6	0.9.6

Optional	Minimum	Recommend
CUDA	11.6	12.2
deepspeed	0.10.0	0.14.0
bitsandbytes	0.39.0	0.43.1
vllm	0.4.3	0.5.0
flash-attn	2.3.0	2.6.3

Hardware Requirement

* estimated

Method	Bits	7B	13B	30B	70B	110B	8x7B	8x22B
Full	AMP	120GB	240GB	600GB	1200GB	2000GB	900GB	2400GB
Full	16	60GB	120GB	300GB	600GB	900GB	400GB	1200GB
Freeze	16	20GB	40GB	80GB	200GB	360GB	160GB	400GB
LoRA/GaLore/BAdam	16	16GB	32GB	64GB	160GB	240GB	120GB	320GB
QLoRA	8	10GB	20GB	40GB	80GB	140GB	60GB	160GB
QLoRA	4	6GB	12GB	24GB	48GB	72GB	30GB	96GB
QLoRA	2	4GB	8GB	16GB	24GB	48GB	18GB	48GB

Getting Started

Installation

Important

Installation is mandatory.

git clone --depth 1 https://github.com/hiyouga/LLaMA-Factory.git
cd LLaMA-Factory
pip install -e ".[torch,metrics]"

Extra dependencies available: torch, torch-npu, metrics, deepspeed, liger-kernel, bitsandbytes, hqq, eetq, gptq, awq, aqlm, vllm, galore, badam, adam-mini, qwen, modelscope, quality

Tip

Use pip install --no-deps -e . to resolve package conflicts.

Data Preparation

Please refer to data/README.md for checking the details about the format of dataset files. You can either use datasets on HuggingFace / ModelScope hub or load the dataset in local disk.

Note

Please update data/dataset_info.json to use your custom dataset.

Quickstart

Use the following 3 commands to run LoRA fine-tuning, inference and merging of the Llama3-8B-Instruct model, respectively.

llamafactory-cli train examples/train_lora/llama3_lora_sft.yaml
llamafactory-cli chat examples/inference/llama3_lora_sft.yaml
llamafactory-cli export examples/merge_lora/llama3_lora_sft.yaml

See examples/README.md for advanced usage (including distributed training).

Tip

Use llamafactory-cli help to show help information.

Use W&B Logger

To use Weights & Biases for logging experimental results, you need to add the following arguments to yaml files.

report_to: wandb
run_name: test_run # optional

Set WANDB_API_KEY to your key when launching training tasks to log in with your W&B account.

Citation

@misc{masala2024vorbecstiromanecsterecipetrain,
      title={"Vorbe\c{s}ti Rom\^ane\c{s}te?" A Recipe to Train Powerful Romanian LLMs with English Instructions}, 
      author={Mihai Masala and Denis C. Ilie-Ablachim and Alexandru Dima and Dragos Corlatescu and Miruna Zavelca and Ovio Olaru and Simina Terian and Andrei Terian and Marius Leordeanu and Horia Velicu and Marius Popescu and Mihai Dascalu and Traian Rebedea},
      year={2024},
      eprint={2406.18266},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2406.18266}, 
}

Acknowledgement

This repo benefits from LLaMA-Factory. We thank them for their wonderful work.