Instructions to use adalaw/MAmmoTH-7B-Mistral-MFT with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use adalaw/MAmmoTH-7B-Mistral-MFT with Transformers:

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

pipe = pipeline("text-generation", model="adalaw/MAmmoTH-7B-Mistral-MFT")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("adalaw/MAmmoTH-7B-Mistral-MFT")
model = AutoModelForCausalLM.from_pretrained("adalaw/MAmmoTH-7B-Mistral-MFT", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use adalaw/MAmmoTH-7B-Mistral-MFT with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "adalaw/MAmmoTH-7B-Mistral-MFT"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "adalaw/MAmmoTH-7B-Mistral-MFT",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/adalaw/MAmmoTH-7B-Mistral-MFT

SGLang

How to use adalaw/MAmmoTH-7B-Mistral-MFT 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 "adalaw/MAmmoTH-7B-Mistral-MFT" \
    --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": "adalaw/MAmmoTH-7B-Mistral-MFT",
		"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 "adalaw/MAmmoTH-7B-Mistral-MFT" \
        --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": "adalaw/MAmmoTH-7B-Mistral-MFT",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use adalaw/MAmmoTH-7B-Mistral-MFT with Docker Model Runner:
```
docker model run hf.co/adalaw/MAmmoTH-7B-Mistral-MFT
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

Introduction

The model is trained with Masked Thought Fine-Tuning (MFT), a simple variant of standard Supervised Fine-Tuning (SFT). You can refer to our code and paper below.

Results

We test it with the Hybrid decoding scripts provided in MAmmoTH.

Model	GSM8K	MATH
adalaw/MAmmoTH-7B-Mistral-MFT	77.10	41.2
TIGER-Lab/MAmmoTH-7B-Mistral-SFT	75.00	40.0

Downloads last month: 6

Safetensors

Model size

7B params

Tensor type

F32

Model tree for adalaw/MAmmoTH-7B-Mistral-MFT

Quantizations

1 model

Dataset used to train adalaw/MAmmoTH-7B-Mistral-MFT

Paper for adalaw/MAmmoTH-7B-Mistral-MFT

Masked Thought: Simply Masking Partial Reasoning Steps Can Improve Mathematical Reasoning Learning of Language Models

Paper • 2403.02178 • Published Mar 4, 2024 • 1

adalaw
/

MAmmoTH-7B-Mistral-MFT