Instructions to use lm-provers/QED-Nano with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use lm-provers/QED-Nano with Transformers:

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

pipe = pipeline("text-generation", model="lm-provers/QED-Nano")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("lm-provers/QED-Nano")
model = AutoModelForCausalLM.from_pretrained("lm-provers/QED-Nano")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Inference
Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use lm-provers/QED-Nano with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "lm-provers/QED-Nano"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "lm-provers/QED-Nano",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/lm-provers/QED-Nano

SGLang

How to use lm-provers/QED-Nano 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 "lm-provers/QED-Nano" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "lm-provers/QED-Nano",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

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 "lm-provers/QED-Nano" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "lm-provers/QED-Nano",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use lm-provers/QED-Nano with Docker Model Runner:
```
docker model run hf.co/lm-provers/QED-Nano
```

Training costs question

by yjernite - opened Feb 13

Discussion

yjernite

Feb 13

Hi! This is a fantastic model, especially for the size 🤗 🐜

Could you provide more information about the costs involved in training to help compare to the larger alternatives? Would be particularly helpful to have overall training tokens, training infra and wall clock, or straight up dollar amounts 💲

Thanks in advance!

lewtun

LM Provers org Feb 16

Hey @yjernite ! I added the training costs for the model here: https://huggingface.co/lm-provers/QED-Nano#software--hardware

Note that QED-Nano is trained on top of QED-Nano-SFT, which has its own training costs here: https://huggingface.co/lm-provers/QED-Nano-SFT#software--hardware

Using ~$3 per H100 hour as a rough estimate, the total cost to train QED-Nano end-to-end is ~ $28k. Not cheap, but still a lot cheaper than training a trillion parameter model :)

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