Instructions to use livadies/MiniMax-M2.7-Ghetto-MoE-8-Experts with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use livadies/MiniMax-M2.7-Ghetto-MoE-8-Experts with Transformers:

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

pipe = pipeline("text-generation", model="livadies/MiniMax-M2.7-Ghetto-MoE-8-Experts", trust_remote_code=True)
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("livadies/MiniMax-M2.7-Ghetto-MoE-8-Experts", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("livadies/MiniMax-M2.7-Ghetto-MoE-8-Experts", trust_remote_code=True, 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 livadies/MiniMax-M2.7-Ghetto-MoE-8-Experts with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "livadies/MiniMax-M2.7-Ghetto-MoE-8-Experts"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "livadies/MiniMax-M2.7-Ghetto-MoE-8-Experts",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/livadies/MiniMax-M2.7-Ghetto-MoE-8-Experts

SGLang

How to use livadies/MiniMax-M2.7-Ghetto-MoE-8-Experts 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 "livadies/MiniMax-M2.7-Ghetto-MoE-8-Experts" \
    --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": "livadies/MiniMax-M2.7-Ghetto-MoE-8-Experts",
		"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 "livadies/MiniMax-M2.7-Ghetto-MoE-8-Experts" \
        --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": "livadies/MiniMax-M2.7-Ghetto-MoE-8-Experts",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use livadies/MiniMax-M2.7-Ghetto-MoE-8-Experts with Docker Model Runner:
```
docker model run hf.co/livadies/MiniMax-M2.7-Ghetto-MoE-8-Experts
```

🩸 MiniMax-M2.7 (229B) -> Ghetto-MoE Edition (8 Experts)

[EN] ⚠️ INFRASTRUCTURE STRESS-TEST KIT

This is a mathematical skeleton (1B parameters) of the giant MiniMax-M2.7 architecture. Created specifically for MLOps engineers to test inference pipelines and routing logic.

Technical Highlights:

8-Expert MoE: Preserved routing logic for Sparse Mixture-of-Experts.
RoPE Fixed: Forced linear rotary embeddings to bypass 'default' KeyError.
8:8 GQA: Symmetrical attention heads for consumer-grade GPU support.

[RU] ⚠️ ИНСТРУМЕНТ ДЛЯ ТЕСТИРОВАНИЯ ИНФРАСТРУКТУРЫ

Это математический скелет (1 млрд параметров) гигантской модели MiniMax-M2.7. Создан для отладки инференса и логики роутинга на обычном железе.

Особенности взлома:

8 Экспертов: Сохранена логика переключения экспертов (routing).
RoPE Patch: Исправлен баг инициализации через принудительный тип linear.
Симметрия GQA: Пропорция голов 8:8 для стабильной работы на старых GPU.

🎧 PROMO DIRECTIVE: RUSSIAN WINTER 26

Слушайте музыку будущего, под которую взламывался этот код:

🔴 YouTube Music: livadies (Search)
🟢 Spotify: livadies on Spotify
🟡 Yandex Music: livadies на Яндекс Музыке

🔥 Главный трек кванта: «RUSSIAN WINTER 26»

Downloads last month: 18

Safetensors

Model size

1B params

Tensor type

F16

Model tree for livadies/MiniMax-M2.7-Ghetto-MoE-8-Experts

Base model

MiniMaxAI/MiniMax-M2.7

Quantized

(112)

this model