Instructions to use raphaeloliveira/Uirapuru-JT-Llama-3.2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use raphaeloliveira/Uirapuru-JT-Llama-3.2 with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="raphaeloliveira/Uirapuru-JT-Llama-3.2")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("raphaeloliveira/Uirapuru-JT-Llama-3.2")
model = AutoModelForCausalLM.from_pretrained("raphaeloliveira/Uirapuru-JT-Llama-3.2", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use raphaeloliveira/Uirapuru-JT-Llama-3.2 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "raphaeloliveira/Uirapuru-JT-Llama-3.2"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "raphaeloliveira/Uirapuru-JT-Llama-3.2",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/raphaeloliveira/Uirapuru-JT-Llama-3.2

SGLang

How to use raphaeloliveira/Uirapuru-JT-Llama-3.2 with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "raphaeloliveira/Uirapuru-JT-Llama-3.2" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "raphaeloliveira/Uirapuru-JT-Llama-3.2",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "raphaeloliveira/Uirapuru-JT-Llama-3.2" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "raphaeloliveira/Uirapuru-JT-Llama-3.2",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use raphaeloliveira/Uirapuru-JT-Llama-3.2 with Docker Model Runner:
```
docker model run hf.co/raphaeloliveira/Uirapuru-JT-Llama-3.2
```

You need to agree to share your contact information to access this model

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

Uirapuru-JT (Llama 3.2)

Uirapuru-JT (Llama 3.2) is a Portuguese-language foundation model specialised in Brazilian Labour Justice. The model was obtained through continued pre-training of Llama 3.2 on a large corpus of Brazilian Labour Court decisions.

Model Details

Model type: Autoregressive Transformer language model
Parameters: Approximately 3.2 billion
Language: Brazilian Portuguese
Domain: Brazilian Labour Justice
Base model: Llama 3.2
Vocabulary size: 128,256 tokens
Developed by: Raphael Souza de Oliveira and Erick Giovani Sperandio Nascimento
License: Apache 2.0

Intended Use

Uirapuru-JT (Llama 3.2) is intended primarily for research and development in Portuguese Natural Language Processing, particularly for applications involving Brazilian Labour Justice.

The model can be used as a foundation for downstream tasks.

Out-of-Scope Use

The model is not intended to:

make autonomous judicial or legal decisions;
replace judges, lawyers, or other legal professionals;
provide authoritative legal advice;
make decisions about individuals without appropriate human oversight;
be used in high-stakes legal applications without independent validation.

Limitations

The model was trained on historical Brazilian Labour Court documents and may reproduce biases, linguistic patterns, and institutional characteristics present in the training data.

The model should be considered a research and decision-support tool rather than an autonomous legal reasoning system.

Training Data

The model was trained on a corpus of Brazilian Labour Court decisions.

The data were anonymised before training. Personally identifiable information was replaced using rule-based and regular-expression procedures.

The training and validation split was 90%/10%.

Training Procedure

Uirapuru-JT (Llama 3.2) was obtained through continued pre-training of Llama 3.2 on the Brazilian Labour Law corpus.

Training was performed using four NVIDIA A100 GPUs with 40 GB of memory each.

Citation

If you use this model in academic research, please cite:

@article{oliveira2026uirapuru,
title = {Uirapuru-JT},
author = {Oliveira, Raphael Souza de and Nascimento, Erick Giovani Sperandio},
year = {2026}
}

Model Card Authors

Raphael Souza de Oliveira Erick Giovani Sperandio Nascimento

Model Card Contact

Raphael Souza de Oliveira: raphael.oliveira@gmail.com
Erick Giovani Sperandio Nascimento: erick.sperandio@gmail.com

Downloads last month: 11

Safetensors

Model size

3B params

Tensor type

BF16

Model tree for raphaeloliveira/Uirapuru-JT-Llama-3.2

Base model

meta-llama/Llama-3.2-3B

Finetuned

(513)

this model