Instructions to use vultr/Meta-Llama-3.1-70B-Instruct-AWQ-INT4-Dequantized-FP32 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use vultr/Meta-Llama-3.1-70B-Instruct-AWQ-INT4-Dequantized-FP32 with Transformers:

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

pipe = pipeline("text-generation", model="vultr/Meta-Llama-3.1-70B-Instruct-AWQ-INT4-Dequantized-FP32")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("vultr/Meta-Llama-3.1-70B-Instruct-AWQ-INT4-Dequantized-FP32")
model = AutoModelForCausalLM.from_pretrained("vultr/Meta-Llama-3.1-70B-Instruct-AWQ-INT4-Dequantized-FP32", 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]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use vultr/Meta-Llama-3.1-70B-Instruct-AWQ-INT4-Dequantized-FP32 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "vultr/Meta-Llama-3.1-70B-Instruct-AWQ-INT4-Dequantized-FP32"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "vultr/Meta-Llama-3.1-70B-Instruct-AWQ-INT4-Dequantized-FP32",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/vultr/Meta-Llama-3.1-70B-Instruct-AWQ-INT4-Dequantized-FP32

SGLang

How to use vultr/Meta-Llama-3.1-70B-Instruct-AWQ-INT4-Dequantized-FP32 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 "vultr/Meta-Llama-3.1-70B-Instruct-AWQ-INT4-Dequantized-FP32" \
    --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": "vultr/Meta-Llama-3.1-70B-Instruct-AWQ-INT4-Dequantized-FP32",
		"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 "vultr/Meta-Llama-3.1-70B-Instruct-AWQ-INT4-Dequantized-FP32" \
        --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": "vultr/Meta-Llama-3.1-70B-Instruct-AWQ-INT4-Dequantized-FP32",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use vultr/Meta-Llama-3.1-70B-Instruct-AWQ-INT4-Dequantized-FP32 with Docker Model Runner:
```
docker model run hf.co/vultr/Meta-Llama-3.1-70B-Instruct-AWQ-INT4-Dequantized-FP32
```

Model Information

The vultr/Meta-Llama-3.1-70B-Instruct-AWQ-INT4-Dequantized-FP32 model is a quantized version Meta-Llama-3.1-70B-Instruct that was dequantized from HuggingFace's AWS Int4 model and requantized and optimized to run on AMD GPUs. It is a drop-in replacement for hugging-quants/Meta-Llama-3.1-70B-Instruct-AWQ-INT4.

Throughput: 68.74 requests/s, 43994.71 total tokens/s, 8798.94 output tokens/s

Model Details

Model Description

Developed by: Meta
Model type: Quantized Large Language Model
Language(s) (NLP): English, German, French, Italian, Portuguese, Hindi, Spanish, and Thai.
License: Llama 3.1
Dequantized From: hugging-quants/Meta-Llama-3.1-70B-Instruct-AWQ-INT4

Compute Infrastructure

Vultr

Hardware

AMD MI300X

Software

ROCm

Model Author

biondizzle

Downloads last month: 8

Safetensors

Model size

71B params

Tensor type

F32

Model tree for vultr/Meta-Llama-3.1-70B-Instruct-AWQ-INT4-Dequantized-FP32

Base model

meta-llama/Llama-3.1-70B

Finetuned

(70)

this model