Instructions to use abacusai/Smaug-Llama-3-70B-Instruct with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use abacusai/Smaug-Llama-3-70B-Instruct with Transformers:

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

pipe = pipeline("text-generation", model="abacusai/Smaug-Llama-3-70B-Instruct")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("abacusai/Smaug-Llama-3-70B-Instruct")
model = AutoModelForCausalLM.from_pretrained("abacusai/Smaug-Llama-3-70B-Instruct")
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 Settings

vLLM

How to use abacusai/Smaug-Llama-3-70B-Instruct with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "abacusai/Smaug-Llama-3-70B-Instruct"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "abacusai/Smaug-Llama-3-70B-Instruct",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/abacusai/Smaug-Llama-3-70B-Instruct

SGLang

How to use abacusai/Smaug-Llama-3-70B-Instruct 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 "abacusai/Smaug-Llama-3-70B-Instruct" \
    --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": "abacusai/Smaug-Llama-3-70B-Instruct",
		"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 "abacusai/Smaug-Llama-3-70B-Instruct" \
        --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": "abacusai/Smaug-Llama-3-70B-Instruct",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use abacusai/Smaug-Llama-3-70B-Instruct with Docker Model Runner:
```
docker model run hf.co/abacusai/Smaug-Llama-3-70B-Instruct
```

Smaug-LLaMa-3-8B-Instruct in the works?

by Joseph717171 - opened May 19, 2024

Discussion

Joseph717171

May 19, 2024

Will you guys make an 8B version too? Please 🙏 😋

ArkaAbacus

May 19, 2024

Actually we already have one! https://huggingface.co/abacusai/Llama-3-Smaug-8B This doesn't have instruct in the name but it is an instruct model. Although we have improved on the recipe since the 8B release so we might update it soon...

Dampfinchen

May 19, 2024

•

edited May 19, 2024

Actually we already have one! https://huggingface.co/abacusai/Llama-3-Smaug-8B This doesn't have instruct in the name but it is an instruct model. Although we have improved on the recipe since the 8B release so we might update it soon...

I think they wanted Smaug 8B trained on LLama 8B Instruct instead of the base model. Models trained on instruct tend to do a lot better, as I'm sure you have noticed with Smaug 70B Instruct.

ArkaAbacus

May 19, 2024

It was trained on Llama 8B Instruct! We just didn't say so in the model card. I've fixed that now.

Joseph717171 changed discussion title from Smaug-LLaMa-8B-Instruct in the works? to Smaug-LLaMa-3-8B-Instruct in the works? May 19, 2024

ManniX-ITA

Jun 6, 2024

@ArkaAbacus
May I ask if you are already planning to update the 8B model with the new recipe and if you have an approximate ETA?
No hurry, I'm just planning to make my own quantization for the ollama library and would like to avoid the new model comes out the day after I'm done with it :P

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