Instructions to use casperhansen/llama-3-70b-instruct-awq with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use casperhansen/llama-3-70b-instruct-awq with Transformers:

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

pipe = pipeline("text-generation", model="casperhansen/llama-3-70b-instruct-awq")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("casperhansen/llama-3-70b-instruct-awq")
model = AutoModelForCausalLM.from_pretrained("casperhansen/llama-3-70b-instruct-awq")
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]:]))

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use casperhansen/llama-3-70b-instruct-awq with vLLM:

Install from pip and serve model

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

Use Docker

docker model run hf.co/casperhansen/llama-3-70b-instruct-awq

SGLang

How to use casperhansen/llama-3-70b-instruct-awq 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 "casperhansen/llama-3-70b-instruct-awq" \
    --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": "casperhansen/llama-3-70b-instruct-awq",
		"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 "casperhansen/llama-3-70b-instruct-awq" \
        --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": "casperhansen/llama-3-70b-instruct-awq",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use casperhansen/llama-3-70b-instruct-awq with Docker Model Runner:
```
docker model run hf.co/casperhansen/llama-3-70b-instruct-awq
```

encountered error when loading model

by zhouzr - opened Apr 20, 2024

Discussion

zhouzr

Apr 20, 2024

vllm 0.3.1

ValueError: Couldn't instantiate the backend tokenizer from one of:
(1) a tokenizers library serialization file,
(2) a slow tokenizer instance to convert or
(3) an equivalent slow tokenizer class to instantiate and convert.
You need to have sentencepiece installed to convert a slow tokenizer to a fast one.

Wizd

Apr 20, 2024

need 0.4.1

PartTimePhilosopher

Apr 20, 2024

but 0.4.1 is not out yet? only 0.4.0.post1?

need 0.4.1

zoltan-fedor

Apr 20, 2024

vllm 0.4.1 is in "pre-release", see https://github.com/vllm-project/vllm/releases

zoltan-fedor

Apr 20, 2024

•

edited Apr 20, 2024

It does seem to run on v0.4.0.post1, as the generation_config.json has the updated eos_token, see https://github.com/vllm-project/vllm/issues/4180#issuecomment-2066187578

casperhansen

Owner Apr 20, 2024

Some versions of vLLM have issues with quantized models. v0.4.0.post1 is the latest that I have confirmed to work personally.

zoltan-fedor

Apr 20, 2024

•

edited Apr 20, 2024

I've just built a docker image for the latest vLLM (for their dev version, so it is v0.4.1.dev, see https://hub.docker.com/r/aiappsref/vllm/tags) and I have just tested it with this quantized model and it seems to be working.
It actually works better than v0.4.0.post1, as that was not finishing its response at the EOS but kept going (that was a known bug fixed in v0.4.1)

stlkcmrd

Apr 26, 2024

I still have it on 0.4.1

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