# Model Card for Model ID *Built with Meta Llama 3* RoLlama3 is a family of pretrained and fine-tuned generative text models for Romanian. This is the repository for the **instruct 8B model**. Links to other models can be found at the bottom of this page. ## Model Details ### Model Description OpenLLM-Ro represents the first open-source effort to build a LLM specialized for Romanian. OpenLLM-Ro developed and publicly releases a collection of Romanian LLMs, both in the form of foundational model and instruct and chat variants. - **Developed by:** OpenLLM-Ro - **Language(s):** Romanian - **License:** cc-by-nc-4.0 - **Finetuned from model:** [Meta-Llama-3-8B](https://huggingface.co/meta-llama/Meta-Llama-3-8B) - **Trained using:** [RoAlpaca](https://huggingface.co/datasets/OpenLLM-Ro/ro_sft_alpaca), [RoAlpacaGPT4](https://huggingface.co/datasets/OpenLLM-Ro/ro_sft_alpaca_gpt4), [RoDolly](https://huggingface.co/datasets/OpenLLM-Ro/ro_sft_dolly), [RoSelfInstruct](https://huggingface.co/datasets/OpenLLM-Ro/ro_sft_selfinstruct_gpt4), [RoNoRobots](https://huggingface.co/datasets/OpenLLM-Ro/ro_sft_norobots), [RoOrca](https://huggingface.co/datasets/OpenLLM-Ro/ro_sft_orca), [RoCamel](https://huggingface.co/datasets/OpenLLM-Ro/ro_sft_camel) ### Model Sources - **Repository:** https://github.com/OpenLLM-Ro/LLaMA-Factory - **Paper:** https://arxiv.org/abs/2406.18266 ## Intended Use ### Intended Use Cases RoLlama3 is intented for research use in Romanian. Base models can be adapted for a variety of natural language tasks while instruction and chat tuned models are intended for assistant-like chat. ### Out-of-Scope Use Use in any manner that violates the license, any applicable laws or regluations, use in languages other than Romanian. ## How to Get Started with the Model Use the code below to get started with the model. ```python from transformers import AutoTokenizer, AutoModelForCausalLM tokenizer = AutoTokenizer.from_pretrained("OpenLLM-Ro/RoLlama3-8b-Instruct-2024-06-28") model = AutoModelForCausalLM.from_pretrained("OpenLLM-Ro/RoLlama3-8b-Instruct-2024-06-28") instruction = "Ce jocuri de societate pot juca cu prietenii mei?" chat = [ {"role": "system", "content": "Ești un asistent folositor, respectuos și onest. Încearcă să ajuți cât mai mult prin informațiile oferite, excluzând răspunsuri toxice, rasiste, sexiste, periculoase și ilegale."}, {"role": "user", "content": instruction}, ] prompt = tokenizer.apply_chat_template(chat, tokenize=False, system_message="") inputs = tokenizer.encode(prompt, add_special_tokens=False, return_tensors="pt") outputs = model.generate(input_ids=inputs, max_new_tokens=128) print(tokenizer.decode(outputs[0])) ``` ## Academic Benchmarks

Model	Average	ARC	MMLU	Winogrande	Hellaswag	GSM8k	TruthfulQA
Llama-3-8B-Instruct	50.62	43.69	52.04	59.33	53.19	43.87	51.59
RoLlama3-8b-Instruct-2024-06-28	50.56	44.70	52.19	67.23	57.69	30.23	51.34
RoLlama3-8b-Instruct-2024-10-09	52.21	47.94	53.50	66.06	59.72	40.16	45.90
RoLlama3-8b-Instruct-DPO-2024-10-09	49.96	46.29	53.29	65.57	58.15	34.77	41.70

## Downstream tasks

	LaRoSeDa				WMT
	Few-shot		Finetuned		Few-shot		Finetuned
Model	Binary (Macro F1)	Multiclass (Macro F1)	Binary (Macro F1)	Multiclass (Macro F1)	EN-RO (Bleu)	RO-EN (Bleu)	EN-RO (Bleu)	RO-EN (Bleu)
Llama-3-8B-Instruct	95.88	56.21	98.53	86.19	18.88	30.98	28.02	40.28
RoLlama3-8b-Instruct-2024-06-28	97.52	67.41	94.15	87.13	24.01	27.36	26.53	40.36
RoLlama3-8b-Instruct-2024-10-09	95.58	61.20	96.46	87.26	22.92	24.28	27.31	40.52
RoLlama3-8b-Instruct-DPO-2024-10-09	97.48	54.00	-	-	22.09	23.00	-	-

	XQuAD				STS
	Few-shot		Finetuned		Few-shot		Finetuned
Model	(EM)	(F1)	(EM)	(F1)	(Spearman)	(Pearson)	(Spearman)	(Pearson)
Llama-3-8B-Instruct	39.47	58.67	67.65	82.77	73.04	72.36	83.49	84.06
RoLlama3-8b-Instruct-2024-06-28	39.43	59.50	44.45	59.76	77.20	77.87	85.80	86.05
RoLlama3-8b-Instruct-2024-10-09	18.89	31.79	50.84	65.18	77.60	76.86	86.70	87.09
RoLlama3-8b-Instruct-DPO-2024-10-09	26.05	42.77	-	-	79.64	79.52	-	-

## MT-Bench

Model	Average	1st turn	2nd turn	Answers in Ro
Llama-3-8B-Instruct	5.96	6.16	5.76	158/160
RoLlama3-8b-Instruct-2024-06-28	5.15	6.03	4.28	160/160
RoLlama3-8b-Instruct-2024-10-09	5.38	6.09	4.67	160/160
RoLlama3-8b-Instruct-DPO-2024-10-09	5.87	6.22	5.49	160/160

## RoCulturaBench

Model	Average	Answers in Ro
Llama-3-8B-Instruct	4.62	100/100
RoLlama3-8b-Instruct-2024-06-28	3.71	100/100
RoLlama3-8b-Instruct-2024-10-09	3.81	100/100
RoLlama3-8b-Instruct-DPO-2024-10-09	4.40	100/100

## RoLlama3 Model Family | Model | Link | |--------------------|:--------:| |*RoLlama3-8b-Instruct-2024-06-28*| [link](https://huggingface.co/OpenLLM-Ro/RoLlama3-8b-Instruct-2024-06-28) | |RoLlama3-8b-Instruct-2024-10-09| [link](https://huggingface.co/OpenLLM-Ro/RoLlama3-8b-Instruct-2024-10-09) | |RoLlama3-8b-Instruct-DPO-2024-10-09| [link](https://huggingface.co/OpenLLM-Ro/RoLlama3-8b-Instruct-DPO-2024-10-09) | ## 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-Dan and Andrei Terian-Dan 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}, } ```