Instructions to use plezan/MiniMax-M2.1-REAP-50-W4A16 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use plezan/MiniMax-M2.1-REAP-50-W4A16 with Transformers:

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

pipe = pipeline("text-generation", model="plezan/MiniMax-M2.1-REAP-50-W4A16", 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("plezan/MiniMax-M2.1-REAP-50-W4A16", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("plezan/MiniMax-M2.1-REAP-50-W4A16", 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 plezan/MiniMax-M2.1-REAP-50-W4A16 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "plezan/MiniMax-M2.1-REAP-50-W4A16"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "plezan/MiniMax-M2.1-REAP-50-W4A16",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/plezan/MiniMax-M2.1-REAP-50-W4A16

SGLang

How to use plezan/MiniMax-M2.1-REAP-50-W4A16 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 "plezan/MiniMax-M2.1-REAP-50-W4A16" \
    --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": "plezan/MiniMax-M2.1-REAP-50-W4A16",
		"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 "plezan/MiniMax-M2.1-REAP-50-W4A16" \
        --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": "plezan/MiniMax-M2.1-REAP-50-W4A16",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use plezan/MiniMax-M2.1-REAP-50-W4A16 with Docker Model Runner:
```
docker model run hf.co/plezan/MiniMax-M2.1-REAP-50-W4A16
```

0xSero/MiniMax-M2.1-REAP-50-W4A16

by itsmebcc - opened Jan 14

Discussion

itsmebcc

Jan 14

Why was 0xSero's version removed from his repo?

nmitchko

Jan 20

I think the original was "broken" aka, it got into loops. This one will too without the right settings overrides

plezan

Owner Jan 22

Yes. He didn't want to have a model on his HF with a looping issue. As this issue wasn't a deal-breaker for my personal use, i thought it could be useful to some.
That's why i re-uploaded it with the looping issues specified on the model card

ttomasone

Feb 13

This model's great. I'm using it on my DGX Spark with minimal issues. Hopefully we get MiniMax 2.5 quants like this soon.

itsmebcc

Feb 13

I think the original was "broken" aka, it got into loops. This one will too without the right settings overrides

I have not had any looping issues with this. What are the suggested overrrides you speak of?

plezan

Owner Feb 13

I didn't have any issues either but 0xSero had some in his tests (link of the tests & results in the model card)

ttomasone

Feb 13

Do you think we'll get a MiniMax 2.5 quant version around the same size? Can't wait to try it.

plezan

Owner Feb 13

If i have some time/money available i might try to reap/quantize 2.5 using the same method as 0xSero, maybe he will do it as well (probably better than me)

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