Instructions to use CompiwerAI/Mtrini-27B-Tellus-Merged with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use CompiwerAI/Mtrini-27B-Tellus-Merged with Transformers:

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

pipe = pipeline("text-generation", model="CompiwerAI/Mtrini-27B-Tellus-Merged")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("CompiwerAI/Mtrini-27B-Tellus-Merged")
model = AutoModelForCausalLM.from_pretrained("CompiwerAI/Mtrini-27B-Tellus-Merged", device_map="auto")
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=256)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use CompiwerAI/Mtrini-27B-Tellus-Merged with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "CompiwerAI/Mtrini-27B-Tellus-Merged"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "CompiwerAI/Mtrini-27B-Tellus-Merged",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/CompiwerAI/Mtrini-27B-Tellus-Merged

SGLang

How to use CompiwerAI/Mtrini-27B-Tellus-Merged 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 "CompiwerAI/Mtrini-27B-Tellus-Merged" \
    --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": "CompiwerAI/Mtrini-27B-Tellus-Merged",
		"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 "CompiwerAI/Mtrini-27B-Tellus-Merged" \
        --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": "CompiwerAI/Mtrini-27B-Tellus-Merged",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use CompiwerAI/Mtrini-27B-Tellus-Merged with Docker Model Runner:
```
docker model run hf.co/CompiwerAI/Mtrini-27B-Tellus-Merged
```

Mtrini-27B-Tellus-Merged

File size: 1,592 Bytes

4cc4f45

---
license: apache-2.0
tags:
- text-generation
- causal-language-model
- merged-model
- qwen
- coding
- mathematics
- reasoning
- darija
- arabic
library_name: transformers
pipeline_tag: text-generation
---

# Mtrini-27B-Tellus-Merged

This repository contains the **merged Transformers model** for Mtrini-27B-Tellus.

The CompiwerAI LoRA adapter has been merged into the compatible base model, producing a standalone model for Transformers-based inference.

> **CompiwerAI — Building AI For Everyone.**

## Model Information

| Property | Value |
|---|---|
| Model | Mtrini-27B-Tellus |
| Type | Merged Transformers model |
| Base | Qwen3.8-27B |
| Parameters | ~27B |
| Context | 4096 tokens |
| Training steps | 700 |
| Training method | QLoRA / LoRA |
| LoRA rank | 32 |
| LoRA alpha | 64 |
| Learning rate | 0.00015 |
| Compute dtype | BF16 |

## Training Mix

- OpenCodeInstruct — 35%
- Magicoder — 10%
- OpenR1-Math — 25%
- Moroccan Darija — 20%
- Aya Arabic / Moroccan Arabic — 10%

Total: **11,200 examples**

## Training Result

- Final loss: **0.4302458722250802**
- Runtime: approximately **3.15 hours**
- Epoch: **1**
- Hardware: **NVIDIA RTX PRO 6000 Blackwell Server Edition**

## Transformers Usage

```python
from transformers import AutoTokenizer, AutoModelForCausalLM

model_id = "CompiwerAI/Mtrini-27B-Tellus-Merged"

tokenizer = AutoTokenizer.from_pretrained(model_id)

model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype="auto",
    device_map="auto",
)
```

For large-model inference, substantial GPU or CPU memory may be required.