Instructions to use marinarosa/minicpm5-1b-vivamais-v0 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use marinarosa/minicpm5-1b-vivamais-v0 with Transformers:

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

pipe = pipeline("text-generation", model="marinarosa/minicpm5-1b-vivamais-v0")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForMultimodalLM

tokenizer = AutoTokenizer.from_pretrained("marinarosa/minicpm5-1b-vivamais-v0")
model = AutoModelForMultimodalLM.from_pretrained("marinarosa/minicpm5-1b-vivamais-v0")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use marinarosa/minicpm5-1b-vivamais-v0 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "marinarosa/minicpm5-1b-vivamais-v0"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "marinarosa/minicpm5-1b-vivamais-v0",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/marinarosa/minicpm5-1b-vivamais-v0

SGLang

How to use marinarosa/minicpm5-1b-vivamais-v0 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 "marinarosa/minicpm5-1b-vivamais-v0" \
    --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": "marinarosa/minicpm5-1b-vivamais-v0",
		"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 "marinarosa/minicpm5-1b-vivamais-v0" \
        --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": "marinarosa/minicpm5-1b-vivamais-v0",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use marinarosa/minicpm5-1b-vivamais-v0 with Docker Model Runner:
```
docker model run hf.co/marinarosa/minicpm5-1b-vivamais-v0
```

minicpm5-1b-vivamais-v0

MiniCPM5-1B text fine-tune for Viva Mais dashboard Q&A.

Training uses a redacted-only mix: public Portuguese instruction data, schema-shaped Viva Mais dashboard examples, and grounding/refusal cases. Raw WhatsApp exports, full transcriptions, and client identifiers are not published.

Acceptance is the repository's Viva Mais QA eval, not generic Portuguese leaderboards.

Viva Mais QA Eval

This v0 checkpoint is candidate 002 from the Viva Mais text SFT pipeline. It was published by user decision because it beat the base model on average score, but it is not marked as fully accepted by the automated gate.

Base model: openbmb/MiniCPM5-1B
Eval run: vivamais_qa_candidate_002
Average score: 0.6588541666666667
Base average score: 0.6197916666666667
Pass rate: 0.5625
Gate passed: false
Leakage failures: 2
Unknown-answer failures: 3

Use this checkpoint as a versioned Viva Mais candidate, not as the final accepted grounded-Q&A model.

Downloads last month: 84

Safetensors

Model size

1B params

Tensor type

BF16

Model tree for marinarosa/minicpm5-1b-vivamais-v0

Base model

openbmb/MiniCPM5-1B

Finetuned

(23)

this model

Finetunes

1 model