Instructions to use benjamin/Gemma2-2B-Distilled-Math with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use benjamin/Gemma2-2B-Distilled-Math with Transformers:

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

pipe = pipeline("text-generation", model="benjamin/Gemma2-2B-Distilled-Math")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("benjamin/Gemma2-2B-Distilled-Math")
model = AutoModelForCausalLM.from_pretrained("benjamin/Gemma2-2B-Distilled-Math", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use benjamin/Gemma2-2B-Distilled-Math with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "benjamin/Gemma2-2B-Distilled-Math"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "benjamin/Gemma2-2B-Distilled-Math",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/benjamin/Gemma2-2B-Distilled-Math

SGLang

How to use benjamin/Gemma2-2B-Distilled-Math 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 "benjamin/Gemma2-2B-Distilled-Math" \
    --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": "benjamin/Gemma2-2B-Distilled-Math",
		"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 "benjamin/Gemma2-2B-Distilled-Math" \
        --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": "benjamin/Gemma2-2B-Distilled-Math",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use benjamin/Gemma2-2B-Distilled-Math with Docker Model Runner:
```
docker model run hf.co/benjamin/Gemma2-2B-Distilled-Math
```

Gemma2-2B-Distilled-Math / README.md

benjamin

Update README.md

66a3e78 verified over 1 year ago

preview code

Raw

History Blame Contribute Delete

1.72 kB

metadata

library_name: transformers
datasets:
  - nvidia/OpenMathInstruct-2
base_model:
  - google/gemma-2-2b-it

Gemma2-2B Distilled from OpenMath2-Llama3.1-8B Model Card

Gemma2-2B distilled from OpenMath2-Llama3.1-8B for math tasks.

This model greatly outperforms the general-purpose Gemma2 instruction-tuning finetune on math tasks.

Model Details

Base Model: Gemma2-2B
Tokenization: Gemma2-2B
Training Methodology: Distillation from OpenMath2-Llama3.1-8B on OpenMathInstruct-2.

Benchmark	Gemma2-2B-Distilled-Math	Original Gemma2-2B-IT
GSM8K (zero-shot)	65.1	6.1
MATH (zero-shot)	52.1	9.9

Model Details

Details on the training methodology are forthcoming.

Use

import torch
from transformers import pipeline

template = "<|start_header_id|>user<|end_header_id|>\n\nSolve the following math problem. Make sure to put the answer (and only answer) inside \boxed{}.\n\n{{problem}}<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n"
problem = "What is the minimum value of $a^2+6a-7$?"

pipe = pipeline(
    "text-generation",
    model="benjamin/Gemma2-2B-Distilled-Math",
    model_kwargs={"torch_dtype": torch.bfloat16},
    eos_token_id=107,
    device_map="auto",
)


outputs = pipe(template.format(problem), max_new_tokens=256)
assistant_response = outputs[0]["generated_text"].strip()
print(assistant_response)