Instructions to use Lasimeri/MiniMax-M2.7-int4-AutoRound with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Lasimeri/MiniMax-M2.7-int4-AutoRound with Transformers:

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

pipe = pipeline("text-generation", model="Lasimeri/MiniMax-M2.7-int4-AutoRound", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Lasimeri/MiniMax-M2.7-int4-AutoRound", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("Lasimeri/MiniMax-M2.7-int4-AutoRound", trust_remote_code=True)
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

vLLM

How to use Lasimeri/MiniMax-M2.7-int4-AutoRound with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Lasimeri/MiniMax-M2.7-int4-AutoRound"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Lasimeri/MiniMax-M2.7-int4-AutoRound",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Lasimeri/MiniMax-M2.7-int4-AutoRound

SGLang

How to use Lasimeri/MiniMax-M2.7-int4-AutoRound 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 "Lasimeri/MiniMax-M2.7-int4-AutoRound" \
    --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": "Lasimeri/MiniMax-M2.7-int4-AutoRound",
		"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 "Lasimeri/MiniMax-M2.7-int4-AutoRound" \
        --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": "Lasimeri/MiniMax-M2.7-int4-AutoRound",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Lasimeri/MiniMax-M2.7-int4-AutoRound with Docker Model Runner:
```
docker model run hf.co/Lasimeri/MiniMax-M2.7-int4-AutoRound
```

OutOfMemory during weights loading (vLLM)

by nephepritou - opened 4 days ago

Discussion

nephepritou

4 days ago

Here is my llama-swap config:

  minimax-m2.7-4bit:
    env:
      - VLLM_LOG_STATS_INTERVAL=5
      - VLLM_MARLIN_USE_ATOMIC_ADD=1
      - CUDA_DEVICE_ORDER=PCI_BUS_ID
      - CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7
      - OMP_NUM_THREADS=4
      - VIRTUAL_ENV=/home/gleb/llm/env_qwen_35
    cmd: |
      /home/gleb/.local/bin/uv run
        -m vllm.entrypoints.openai.api_server
        --model /mnt/data/llm-data/models/Lasimeri/MiniMax-M2.7-int4-AutoRound
        #--quantization gptq_marlin
        --served-model-name "minimax-m2.7-4bit"
        --port ${PORT}
        -tp 8
        #--performance-mode interactivity
        --enable-sleep-mode
        --max-num-batched-tokens 8192
        --enable-prefix-caching
        --enable-chunked-prefill
        #--kv-offloading-size 32
        #--disable-hybrid-kv-cache-manager
        --max-model-len auto
        --gpu-memory-utilization 0.9
        --max-num-seqs 2
        --attention-backend flashinfer
        #--kv-cache-dtype float16
        --dtype half
        --reasoning-parser minimax_m2
        --enable-auto-tool-choice
        --tool-call-parser minimax_m2
        #--load-format instanttensor
        --trust-remote-code
        #--enable-expert-parallel
        #--speculative-config '{"method":"suffix","num_speculative_tokens":16}'

With expert parallel it loading and running fine, but slow. Without expert parallel it crashes with OOM during weights loading. Running on 8x3090.

nephepritou changed discussion title from OutOfMemory to OutOfMemory during weights loading 4 days ago

nephepritou changed discussion title from OutOfMemory during weights loading to OutOfMemory during weights loading (vLLM) 4 days ago

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