Instructions to use Inferact/MiniMax-M3-EAGLE3-GQA with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Inferact/MiniMax-M3-EAGLE3-GQA with Transformers:

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

pipe = pipeline("text-generation", model="Inferact/MiniMax-M3-EAGLE3-GQA")

# Load model directly
from transformers import AutoTokenizer, LlamaForCausalLMEagle3

tokenizer = AutoTokenizer.from_pretrained("Inferact/MiniMax-M3-EAGLE3-GQA")
model = LlamaForCausalLMEagle3.from_pretrained("Inferact/MiniMax-M3-EAGLE3-GQA", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use Inferact/MiniMax-M3-EAGLE3-GQA with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Inferact/MiniMax-M3-EAGLE3-GQA"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Inferact/MiniMax-M3-EAGLE3-GQA",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/Inferact/MiniMax-M3-EAGLE3-GQA

SGLang

How to use Inferact/MiniMax-M3-EAGLE3-GQA 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 "Inferact/MiniMax-M3-EAGLE3-GQA" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Inferact/MiniMax-M3-EAGLE3-GQA",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

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 "Inferact/MiniMax-M3-EAGLE3-GQA" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Inferact/MiniMax-M3-EAGLE3-GQA",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use Inferact/MiniMax-M3-EAGLE3-GQA with Docker Model Runner:
```
docker model run hf.co/Inferact/MiniMax-M3-EAGLE3-GQA
```

ZixiQi commited on 7 days ago

Commit

c7b1d3b

verified ·

1 Parent(s): 4a97a1b

Trim Performance section to MT-Bench + SPEED-Bench acceptance; drop MHA comparison

Browse files

Files changed (1) hide show

README.md +6 -12

README.md CHANGED Viewed

@@ -25,15 +25,9 @@ The draft is a 1-layer dense Llama (`LlamaForCausalLMEagle3`) on MiniMax-M3's `h
 ## Performance
-Mean accepted length measured end-to-end against `MiniMaxAI/MiniMax-M3-MXFP8` served with vLLM at `tensor-parallel-size=4`, `num_speculative_tokens=3`, `--enforce-eager`, greedy draft sampling (`topk=1`). The MHA version is shown for reference under identical conditions.
-| Category | Dataset | n | GQA (this model) | MHA reference |
-|---|---|---:|---:|---:|
-| Dialogue | [MT-Bench](https://github.com/lm-sys/FastChat/tree/main/fastchat/llm_judge) | 80 | 2.689 | 2.698 |
-| Math | [GSM8K](https://github.com/openai/grade-school-math) | 200 | 3.500 | 3.518 |
-| Code | [HumanEval](https://huggingface.co/datasets/openai/openai_humaneval) | 164 | 3.489 | 3.499 |
-| Math | [MATH500](https://huggingface.co/datasets/HuggingFaceH4/MATH-500) | 500 | 3.510 | 3.517 |
-| Math | [AIME](https://huggingface.co/datasets/Maxwell-Jia/AIME_2024) | 30 | 3.282 | 3.291 |
-| | **Mean** | | **3.294** | 3.305 |
-GQA matches the MHA version's acceptance (**~99.7%** of mean accepted length) while using **16× fewer** key/value heads.

 ## Performance
+Mean accepted length and draft accept rate measured end-to-end against `MiniMaxAI/MiniMax-M3-MXFP8` served with vLLM at `tensor-parallel-size=4`, `num_speculative_tokens=3`, greedy sampling (`temperature=0`, `top_p=1.0`), `max-concurrency=16`.
+| Dataset | n | Mean accepted length | Draft accept rate |
+|---|---:|---:|---:|
+| MT-Bench | 64 | 2.668 | 55.62% |
+| SPEED-Bench (qualitative) | 64 | 2.561 | 52.04% |