Instructions to use Sreenington/Phi-3-mini-4k-instruct-AWQ with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Sreenington/Phi-3-mini-4k-instruct-AWQ with Transformers:

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

pipe = pipeline("text-generation", model="Sreenington/Phi-3-mini-4k-instruct-AWQ")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Sreenington/Phi-3-mini-4k-instruct-AWQ")
model = AutoModelForCausalLM.from_pretrained("Sreenington/Phi-3-mini-4k-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 Sreenington/Phi-3-mini-4k-instruct-AWQ with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Sreenington/Phi-3-mini-4k-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": "Sreenington/Phi-3-mini-4k-instruct-AWQ",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Sreenington/Phi-3-mini-4k-instruct-AWQ

SGLang

How to use Sreenington/Phi-3-mini-4k-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 "Sreenington/Phi-3-mini-4k-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": "Sreenington/Phi-3-mini-4k-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 "Sreenington/Phi-3-mini-4k-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": "Sreenington/Phi-3-mini-4k-instruct-AWQ",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Sreenington/Phi-3-mini-4k-instruct-AWQ with Docker Model Runner:
```
docker model run hf.co/Sreenington/Phi-3-mini-4k-instruct-AWQ
```

Quantization approach

by Bhawanaprasain10 - opened Jun 24, 2024

Discussion

Bhawanaprasain10

Jun 24, 2024

Hey @Sreenington how did you get the AWQ version for phi3 model. Can you share the approach? I am trying to do AWQ for phi3 with the approach mentioned here and I get an error that says phi3 is not supported for AWQ. However on the docs from vllm I can see phi3 on the supported list. That's sort of contradictory. Can you help?

Sreenington

Owner Jul 5, 2024

Hey man, are you sure? AWQ supports Phi3, try to follow their instructions on their GitHub page (https://github.com/casper-hansen/AutoAWQ/blob/main/examples/quantize.py).
Also are you sure that you ain't facing an obscure CUDA compatibility error? Make sure you have a GPU with 24GB VRAM and 7.5+ CUDA Compute Capability (I used an A10G).

bee133

Jul 16, 2024

The latest release (0.2.5) doesn't have kernel for phi3 yet, the latest branch supports Phi3 though. You should download the repo and directly access the classes.

Bhawanaprasain10

Jul 16, 2024

@Sreenington I am running quantization on a 24 GB GPU and it's not CUDA compatibility issue.

av-codes

Jul 27, 2024

Why does it specify Mistal model in the config.json?

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