Instructions to use Q-bert/MetaMath-Cybertron-Starling with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Q-bert/MetaMath-Cybertron-Starling with Transformers:

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

pipe = pipeline("text-generation", model="Q-bert/MetaMath-Cybertron-Starling")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Q-bert/MetaMath-Cybertron-Starling")
model = AutoModelForCausalLM.from_pretrained("Q-bert/MetaMath-Cybertron-Starling")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use Q-bert/MetaMath-Cybertron-Starling with vLLM:

Install from pip and serve model

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

Use Docker

docker model run hf.co/Q-bert/MetaMath-Cybertron-Starling

SGLang

How to use Q-bert/MetaMath-Cybertron-Starling 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 "Q-bert/MetaMath-Cybertron-Starling" \
    --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": "Q-bert/MetaMath-Cybertron-Starling",
		"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 "Q-bert/MetaMath-Cybertron-Starling" \
        --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": "Q-bert/MetaMath-Cybertron-Starling",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use Q-bert/MetaMath-Cybertron-Starling with Docker Model Runner:
```
docker model run hf.co/Q-bert/MetaMath-Cybertron-Starling
```

This was amazing! The best Mistral 7b I've tested.

by deleted - opened Dec 9, 2023

Discussion

deleted

Dec 9, 2023

Each of the over 20 Mistrals I've tested had notable blind spots, and all the Mergers often had odd outputs.

This one not only scored higher than any other in my person testing, but didn't score low in any area.

It kinda makes sense since Starling is not very bright, but is unusually good at giving human aligned responses, while the other two are unusually smart, good at math... So it makes sense they would compliment each other well.

Again, great job!

TravelingMan

Dec 9, 2023

This really is a great model, one of the best I've used at 7B. Maybe the very best. Seems like it can handle anything, within reason of course.

Q-bert

Owner Dec 10, 2023

Each of the over 20 Mistrals I've tested had notable blind spots, and all the Mergers often had odd outputs.

This one not only scored higher than any other in my person testing, but didn't score low in any area.

It kinda makes sense since Starling is not very bright, but is unusually good at giving human aligned responses, while the other two are unusually smart, good at math... So it makes sense they would compliment each other well.

Again, great job!

This really is a great model, one of the best I've used at 7B. Maybe the very best. Seems like it can handle anything, within reason of course.

Thank you very much for your interest and compliments. I'm glad I could meet your expectations.

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