SaoLa-3B-Instruct

Model Description

SaoLa-3B-Instruct is a Vietnamese Large Language Model built on Qwen2.5-3B-Instruct, with multi-stage continual pretraining and instruction tuning to enhance Vietnamese performance across diverse tasks.

Quickstart

Here provides a code snippet with apply_chat_template to show you how to load the tokenizer and model and how to generate contents.

from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "uonlp2/SaoLa-3B-Instruct"

model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype="auto",
    device_map="auto"
)
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "Giới thiệu về Sao La"
messages = [
    {"role": "system", "content": "Bạn là một trợ lý hữu ích."},
    {"role": "user", "content": prompt}
]
text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True
)
model_inputs = tokenizer([text], return_tensors="pt").to(model.device)

generated_ids = model.generate(
    **model_inputs,
    max_new_tokens=128
)
generated_ids = [
    output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)
]

response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]

Citation

If you find our work helpful, feel free to give us a cite.

@article{saola-3b-instruct,
      title={SaoLa}, 
      author={Thuat Nguyen, Chien Van Nguyen, Linh Ngo Van, Thien Nguyen},
      year={2025}
}