Instructions to use marcosremar2/tucano2-1p5b-ptbr-roleplay-lora-9k with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use marcosremar2/tucano2-1p5b-ptbr-roleplay-lora-9k with PEFT:

from peft import PeftModel
from transformers import AutoModelForCausalLM

base_model = AutoModelForCausalLM.from_pretrained("Polygl0t/Tucano2-qwen-1.5B-Instruct")
model = PeftModel.from_pretrained(base_model, "marcosremar2/tucano2-1p5b-ptbr-roleplay-lora-9k")

Transformers

How to use marcosremar2/tucano2-1p5b-ptbr-roleplay-lora-9k with Transformers:

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

pipe = pipeline("text-generation", model="marcosremar2/tucano2-1p5b-ptbr-roleplay-lora-9k")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("marcosremar2/tucano2-1p5b-ptbr-roleplay-lora-9k", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use marcosremar2/tucano2-1p5b-ptbr-roleplay-lora-9k with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "marcosremar2/tucano2-1p5b-ptbr-roleplay-lora-9k"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "marcosremar2/tucano2-1p5b-ptbr-roleplay-lora-9k",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/marcosremar2/tucano2-1p5b-ptbr-roleplay-lora-9k

SGLang

How to use marcosremar2/tucano2-1p5b-ptbr-roleplay-lora-9k 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 "marcosremar2/tucano2-1p5b-ptbr-roleplay-lora-9k" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "marcosremar2/tucano2-1p5b-ptbr-roleplay-lora-9k",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

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 "marcosremar2/tucano2-1p5b-ptbr-roleplay-lora-9k" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "marcosremar2/tucano2-1p5b-ptbr-roleplay-lora-9k",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use marcosremar2/tucano2-1p5b-ptbr-roleplay-lora-9k with Docker Model Runner:
```
docker model run hf.co/marcosremar2/tucano2-1p5b-ptbr-roleplay-lora-9k
```

Tucano2 1.5B PT-BR Roleplay LoRA 9k

LoRA adapter for Polygl0t/Tucano2-qwen-1.5B-Instruct, fine-tuned for short Brazilian Portuguese role-play conversations.

This is an adapter-only upload. Load it with the original base model.

Intended Use

Simulate a simple character in daily-life conversations.
Maintain the assigned role.
Answer in short, natural Brazilian Portuguese.
Support voice-agent style interactions where another system handles speech input/output.

This adapter is not intended to be a grammar teacher or a precise linguistic correction model.

Training

Base model: Polygl0t/Tucano2-qwen-1.5B-Instruct
Method: LoRA SFT
Trainable parameters: 17,432,576
Dataset: 1,811 approved synthetic PT-BR role-play conversations
SFT examples: 9,055 prompt/completion turns
Train split: 8,512 examples
Validation split: 543 examples
Epochs: 1
Max sequence length: 768
Learning rate: 1e-4
LoRA rank: 16
LoRA alpha: 32
LoRA dropout: 0.05
Hardware: RTX 3090 24 GB
Training runtime: about 22 minutes

Evaluation

Manual benchmark with 30 turns across daily-life role-play scenarios:

Model	Good	Partial	Bad	Peak CUDA VRAM
1.5B Instruct original FP16	9/30	16/30	5/30	n/a
1.5B Instruct + LoRA FP16	15/30	9/30	6/30	2.9 GB
1.5B Instruct original 4-bit NF4	5/30	13/30	12/30	1.44 GB
1.5B Instruct + LoRA 4-bit NF4	16/30	11/30	3/30	1.44 GB
3.7B Instruct original 4-bit NF4	13/30	12/30	5/30	3.6 GB

The LoRA improved short PT-BR role-play behavior substantially, especially after 4-bit loading. Remaining weak points are precise grammar explanations and some direction-giving scenes.

Loading Example

from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel

base_model = "Polygl0t/Tucano2-qwen-1.5B-Instruct"
adapter = "YOUR_USERNAME/tucano2-1p5b-ptbr-roleplay-lora-9k"

tokenizer = AutoTokenizer.from_pretrained(base_model)
model = AutoModelForCausalLM.from_pretrained(base_model, device_map="auto")
model = PeftModel.from_pretrained(model, adapter)
model.eval()

Notes

This adapter does not generate Mimi/audio tokens by itself.
For audio output, connect it to a separate Talker/adaptor trained to predict Mimi codes.
The repo is intended to be private while the dataset and product direction are still experimental.

Downloads last month: 6

Model tree for marcosremar2/tucano2-1p5b-ptbr-roleplay-lora-9k

Base model

Qwen/Qwen3-1.7B-Base

Finetuned

Polygl0t/Tucano2-qwen-1.5B-Base

Finetuned

Polygl0t/Tucano2-qwen-1.5B-Instruct

Adapter

(1)

this model