Instructions to use nvidia/Llama3-ChatQA-1.5-70B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use nvidia/Llama3-ChatQA-1.5-70B with Transformers:

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

pipe = pipeline("text-generation", model="nvidia/Llama3-ChatQA-1.5-70B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("nvidia/Llama3-ChatQA-1.5-70B")
model = AutoModelForCausalLM.from_pretrained("nvidia/Llama3-ChatQA-1.5-70B", device_map="auto")
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 nvidia/Llama3-ChatQA-1.5-70B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "nvidia/Llama3-ChatQA-1.5-70B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "nvidia/Llama3-ChatQA-1.5-70B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/nvidia/Llama3-ChatQA-1.5-70B

SGLang

How to use nvidia/Llama3-ChatQA-1.5-70B 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 "nvidia/Llama3-ChatQA-1.5-70B" \
    --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": "nvidia/Llama3-ChatQA-1.5-70B",
		"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 "nvidia/Llama3-ChatQA-1.5-70B" \
        --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": "nvidia/Llama3-ChatQA-1.5-70B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use nvidia/Llama3-ChatQA-1.5-70B with Docker Model Runner:
```
docker model run hf.co/nvidia/Llama3-ChatQA-1.5-70B
```

About tokens used in this model.

by icoicqico - opened May 8, 2024

Discussion

icoicqico

May 8, 2024

•

edited May 8, 2024

Hello, I would like to know what kind of token format is used to train this model, is it the Llama3 token format? Did you train this with any instruction prompt and padding tokens like <|start_header_id|>system<|end_header_id|>, <|eot_id|>? I would like to know because this may help me to desgin the prompt when fine tuning or using the model, thanks!

zihanliu

NVIDIA org May 8, 2024

Hi,
We do not use the special tokens like <|start_header_id|>, <|end_header_id|>, and <|eot_id|>. You can pretty much follow the format in the sample code we provide. that's the format we use to train our model. In addition, you can also refer to https://huggingface.co/datasets/nvidia/ChatQA-Training-Data, where we provide the training recipe.

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