Instructions to use Unbabel/M-Prometheus-14B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Unbabel/M-Prometheus-14B with Transformers:

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

pipe = pipeline("text-generation", model="Unbabel/M-Prometheus-14B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Unbabel/M-Prometheus-14B")
model = AutoModelForCausalLM.from_pretrained("Unbabel/M-Prometheus-14B")
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]:]))

Inference
Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use Unbabel/M-Prometheus-14B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Unbabel/M-Prometheus-14B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Unbabel/M-Prometheus-14B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Unbabel/M-Prometheus-14B

SGLang

How to use Unbabel/M-Prometheus-14B 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 "Unbabel/M-Prometheus-14B" \
    --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": "Unbabel/M-Prometheus-14B",
		"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 "Unbabel/M-Prometheus-14B" \
        --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": "Unbabel/M-Prometheus-14B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Unbabel/M-Prometheus-14B with Docker Model Runner:
```
docker model run hf.co/Unbabel/M-Prometheus-14B
```

Unexpected outputs

by deleted - opened Jul 24, 2025

Discussion

deleted

Jul 24, 2025

Hello! I’ve been experiencing some issues while using the M-Prometheus models for Direct Assessment on French texts. The 3B version works perfectly, but I’ve encountered many instances of unexpected behaviors with the 7B and 14B models.

For example, here are some outputs I got from the 14B model:

" Feedback: <|im_end|>
<|im_start|>.AFDMETHODIMP CMyClass::MyMethod() {
// Your implementation here
}<|im_end|> "

I prompted the models using Qwen’s chat template, with the prompt you specified. Do you have any idea why I’m getting these unexpected outputs? Thanks in advance!

jmprcp

Jul 24, 2025

Hello,
I've never encountered this sort of issue. Are you using vllm for inference?

deleted

Aug 13, 2025

Hello,
I used the Transformers library for inference, not vllm. I was able to find the cause of my problem: I was preparing the inputs and decoding the outputs similarly to how it is done for Prometheus-2, which is a bit different from the code currently shown on the M-Prometheus model card. Now, all three models work perfectly!

deleted changed discussion status to closed Aug 13, 2025

jmprcp

Aug 13, 2025

Great! Closing the issue, then.

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment