Instructions to use nasos10/MuToR-gemma-2B-1M_MATH-dmax_3_a_02 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use nasos10/MuToR-gemma-2B-1M_MATH-dmax_3_a_02 with Transformers:

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

pipe = pipeline("text-generation", model="nasos10/MuToR-gemma-2B-1M_MATH-dmax_3_a_02")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("nasos10/MuToR-gemma-2B-1M_MATH-dmax_3_a_02")
model = AutoModelForCausalLM.from_pretrained("nasos10/MuToR-gemma-2B-1M_MATH-dmax_3_a_02")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use nasos10/MuToR-gemma-2B-1M_MATH-dmax_3_a_02 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "nasos10/MuToR-gemma-2B-1M_MATH-dmax_3_a_02"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "nasos10/MuToR-gemma-2B-1M_MATH-dmax_3_a_02",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/nasos10/MuToR-gemma-2B-1M_MATH-dmax_3_a_02

SGLang

How to use nasos10/MuToR-gemma-2B-1M_MATH-dmax_3_a_02 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 "nasos10/MuToR-gemma-2B-1M_MATH-dmax_3_a_02" \
    --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": "nasos10/MuToR-gemma-2B-1M_MATH-dmax_3_a_02",
		"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 "nasos10/MuToR-gemma-2B-1M_MATH-dmax_3_a_02" \
        --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": "nasos10/MuToR-gemma-2B-1M_MATH-dmax_3_a_02",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use nasos10/MuToR-gemma-2B-1M_MATH-dmax_3_a_02 with Docker Model Runner:
```
docker model run hf.co/nasos10/MuToR-gemma-2B-1M_MATH-dmax_3_a_02
```

MuToR: Multi-Token prediction with Registers

Arxiv: https://arxiv.org/abs/2505.10518

TL;DR: MuToR is a simple, plug-and-play approach for multi-token prediction. It leverages dummy register tokens to predict multiple targets in the future, enriching the supervisory signal and improving performance across diverse settings and modalities. The register tokens are discarded on inference, leaving generation speed unchanged.

Model Description

This model is a finetuned version of Gemma-2B. It was finetuned using the MuToR method for 5 epochs on the 1M-MATH training split. Please refer to our code for guidelines on how to use the models to reproduce our results.

Downloads last month: 2

Safetensors

Model size

3B params

Tensor type

BF16

Model tree for nasos10/MuToR-gemma-2B-1M_MATH-dmax_3_a_02

Base model

google/gemma-2b

Finetuned

(286)

this model

Dataset used to train nasos10/MuToR-gemma-2B-1M_MATH-dmax_3_a_02

Paper for nasos10/MuToR-gemma-2B-1M_MATH-dmax_3_a_02

Multi-Token Prediction Needs Registers

Paper • 2505.10518 • Published May 15, 2025 • 14