ro base_model:
meta-llama/Meta-Llama-3-8B datasets:
OpenLLM-Ro/ro_sft_alpaca
OpenLLM-Ro/ro_sft_alpaca_gpt4
OpenLLM-Ro/ro_sft_dolly
OpenLLM-Ro/ro_sft_selfinstruct_gpt4
OpenLLM-Ro/ro_sft_norobots
OpenLLM-Ro/ro_sft_orca
OpenLLM-Ro/ro_sft_camel model-index:
- name: OpenLLM-Ro/RoLlama3-8b-Instruct-2024-06-28 results:
  - task: type: text-generation dataset: name: RoMT-Bench type: RoMT-Bench metrics:
    - name: Score type: Score value: 5.15
  - task: type: text-generation dataset: name: RoCulturaBench type: RoCulturaBench metrics:
    - name: Score type: Score value: 3.71
  - task: type: text-generation dataset: name: Romanian_Academic_Benchmarks type: Romanian_Academic_Benchmarks metrics:
    - name: Average accuracy type: accuracy value: 50.56
  - task: type: text-generation dataset: name: OpenLLM-Ro/ro_arc_challenge type: OpenLLM-Ro/ro_arc_challenge metrics:
    - name: Average accuracy type: accuracy value: 44.70
  - task: type: text-generation dataset: name: OpenLLM-Ro/ro_mmlu type: OpenLLM-Ro/ro_mmlu metrics:
    - name: Average accuracy type: accuracy value: 52.19
  - task: type: text-generation dataset: name: OpenLLM-Ro/ro_winogrande type: OpenLLM-Ro/ro_winogrande metrics:
    - name: Average accuracy type: accuracy value: 67.23
  - task: type: text-generation dataset: name: OpenLLM-Ro/ro_hellaswag type: OpenLLM-Ro/ro_hellaswag metrics:
    - name: Average accuracy type: accuracy value: 57.69
  - task: type: text-generation dataset: name: OpenLLM-Ro/ro_gsm8k type: OpenLLM-Ro/ro_gsm8k metrics:
    - name: Average accuracy type: accuracy value: 30.23
  - task: type: text-generation dataset: name: OpenLLM-Ro/ro_truthfulqa type: OpenLLM-Ro/ro_truthfulqa metrics:
    - name: Average accuracy type: accuracy value: 51.34
  - task: type: text-generation dataset: name: LaRoSeDa_binary type: LaRoSeDa_binary metrics:
    - name: Average macro-f1 type: macro-f1 value: 97.52
  - task: type: text-generation dataset: name: LaRoSeDa_multiclass type: LaRoSeDa_multiclass metrics:
    - name: Average macro-f1 type: macro-f1 value: 67.41
  - task: type: text-generation dataset: name: LaRoSeDa_binary_finetuned type: LaRoSeDa_binary_finetuned metrics:
    - name: Average macro-f1 type: macro-f1 value: 94.15
  - task: type: text-generation dataset: name: LaRoSeDa_multiclass_finetuned type: LaRoSeDa_multiclass_finetuned metrics:
    - name: Average macro-f1 type: macro-f1 value: 87.13
  - task: type: text-generation dataset: name: WMT_EN-RO type: WMT_EN-RO metrics:
    - name: Average bleu type: bleu value: 24.01
  - task: type: text-generation dataset: name: WMT_RO-EN type: WMT_RO-EN metrics:
    - name: Average bleu type: bleu value: 27.36
  - task: type: text-generation dataset: name: WMT_EN-RO_finetuned type: WMT_EN-RO_finetuned metrics:
    - name: Average bleu type: bleu value: 26.53
  - task: type: text-generation dataset: name: WMT_RO-EN_finetuned type: WMT_RO-EN_finetuned metrics:
    - name: Average bleu type: bleu value: 40.36
  - task: type: text-generation dataset: name: XQuAD type: XQuAD metrics:
    - name: Average exact_match type: exact_match value: 39.43
  - task: type: text-generation dataset: name: XQuAD type: XQuAD metrics:
    - name: Average f1 type: f1 value: 59.50
  - task: type: text-generation dataset: name: XQuAD_finetuned type: XQuAD_finetuned metrics:
    - name: Average exact_match type: exact_match value: 44.45
  - task: type: text-generation dataset: name: XQuAD_finetuned type: XQuAD_finetuned metrics:
    - name: Average f1 type: f1 value: 59.76
  - task: type: text-generation dataset: name: STS type: STS metrics:
    - name: Average spearman type: spearman value: 77.20
  - task: type: text-generation dataset: name: STS type: STS metrics:
    - name: Average pearson type: pearson value: 77.87
  - task: type: text-generation dataset: name: STS_finetuned type: STS_finetuned metrics:
    - name: Average spearman type: spearman value: 85.80
  - task: type: text-generation dataset: name: STS_finetuned type: STS_finetuned metrics:
    - name: Average pearson type: pearson value: 86.05
  - task: type: text-generation dataset: name: RoMT-Bench type: RoMT-Bench metrics:
    - name: First turn type: Score value: 6.03
    - name: Second turn type: Score value: 4.28
  - task: type: text-generation dataset: name: OpenLLM-Ro/ro_arc_challenge type: OpenLLM-Ro/ro_arc_challenge metrics:
    - name: 0-shot type: accuracy value: 41.90
    - name: 1-shot type: accuracy value: 44.30
    - name: 3-shot type: accuracy value: 44.56
    - name: 5-shot type: accuracy value: 45.50
    - name: 10-shot type: accuracy value: 46.10
    - name: 25-shot type: accuracy value: 45.84
  - task: type: text-generation dataset: name: OpenLLM-Ro/ro_mmlu type: OpenLLM-Ro/ro_mmlu metrics:
    - name: 0-shot type: accuracy value: 50.85
    - name: 1-shot type: accuracy value: 51.24
    - name: 3-shot type: accuracy value: 53.30
    - name: 5-shot type: accuracy value: 53.39
  - task: type: text-generation dataset: name: OpenLLM-Ro/ro_winogrande type: OpenLLM-Ro/ro_winogrande metrics:
    - name: 0-shot type: accuracy value: 65.19
    - name: 1-shot type: accuracy value: 66.54
    - name: 3-shot type: accuracy value: 67.88
    - name: 5-shot type: accuracy value: 69.30
  - task: type: text-generation dataset: name: OpenLLM-Ro/ro_hellaswag type: OpenLLM-Ro/ro_hellaswag metrics:
    - name: 0-shot type: accuracy value: 56.12
    - name: 1-shot type: accuracy value: 57.37
    - name: 3-shot type: accuracy value: 57.92
    - name: 5-shot type: accuracy value: 58.18
    - name: 10-shot type: accuracy value: 58.85
  - task: type: text-generation dataset: name: OpenLLM-Ro/ro_gsm8k type: OpenLLM-Ro/ro_gsm8k metrics:
    - name: 1-shot type: accuracy value: 29.42
    - name: 3-shot type: accuracy value: 30.02
    - name: 5-shot type: accuracy value: 31.24
  - task: type: text-generation dataset: name: LaRoSeDa_binary type: LaRoSeDa_binary metrics:
    - name: 0-shot type: macro-f1 value: 97.43
    - name: 1-shot type: macro-f1 value: 96.60
    - name: 3-shot type: macro-f1 value: 97.90
    - name: 5-shot type: macro-f1 value: 98.13
  - task: type: text-generation dataset: name: LaRoSeDa_multiclass type: LaRoSeDa_multiclass metrics:
    - name: 0-shot type: macro-f1 value: 63.77
    - name: 1-shot type: macro-f1 value: 68.91
    - name: 3-shot type: macro-f1 value: 66.36
    - name: 5-shot type: macro-f1 value: 70.61
  - task: type: text-generation dataset: name: WMT_EN-RO type: WMT_EN-RO metrics:
    - name: 0-shot type: bleu value: 6.92
    - name: 1-shot type: bleu value: 29.33
    - name: 3-shot type: bleu value: 29.79
    - name: 5-shot type: bleu value: 30.02
  - task: type: text-generation dataset: name: WMT_RO-EN type: WMT_RO-EN metrics:
    - name: 0-shot type: bleu value: 4.50
    - name: 1-shot type: bleu value: 30.30
    - name: 3-shot type: bleu value: 36.96
    - name: 5-shot type: bleu value: 37.70
  - task: type: text-generation dataset: name: XQuAD_EM type: XQuAD_EM metrics:
    - name: 0-shot type: exact_match value: 4.45
    - name: 1-shot type: exact_match value: 48.24
    - name: 3-shot type: exact_match value: 52.03
    - name: 5-shot type: exact_match value: 53.03
  - task: type: text-generation dataset: name: XQuAD_F1 type: XQuAD_F1 metrics:
    - name: 0-shot type: f1 value: 26.08
    - name: 1-shot type: f1 value: 68.40
    - name: 3-shot type: f1 value: 71.92
    - name: 5-shot type: f1 value: 71.60
  - task: type: text-generation dataset: name: STS_Spearman type: STS_Spearman metrics:
    - name: 1-shot type: spearman value: 77.76
    - name: 3-shot type: spearman value: 76.72
    - name: 5-shot type: spearman value: 77.12
  - task: type: text-generation dataset: name: STS_Pearson type: STS_Pearson metrics:
    - name: 1-shot type: pearson value: 77.83
    - name: 3-shot type: pearson value: 77.64
    - name: 5-shot type: pearson value: 78.13

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
Trained using: RoAlpaca, RoAlpacaGPT4, RoDolly, RoSelfInstruct, RoNoRobots, RoOrca, RoCamel

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
RoLlama3-8b-Instruct-2024-10-09	link
RoLlama3-8b-Instruct-DPO-2024-10-09	link

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}, 
}

RichardErkhov/OpenLLM-Ro_-_RoLlama3-8b-Instruct-2024-06-28-8bits

Original model description: