Instructions to use microsoft/Phi-3-small-128k-instruct with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use microsoft/Phi-3-small-128k-instruct with Transformers:

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

pipe = pipeline("text-generation", model="microsoft/Phi-3-small-128k-instruct", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("microsoft/Phi-3-small-128k-instruct", trust_remote_code=True, dtype="auto")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use microsoft/Phi-3-small-128k-instruct with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "microsoft/Phi-3-small-128k-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": "microsoft/Phi-3-small-128k-instruct",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/microsoft/Phi-3-small-128k-instruct

SGLang

How to use microsoft/Phi-3-small-128k-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 "microsoft/Phi-3-small-128k-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": "microsoft/Phi-3-small-128k-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 "microsoft/Phi-3-small-128k-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": "microsoft/Phi-3-small-128k-instruct",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use microsoft/Phi-3-small-128k-instruct with Docker Model Runner:
```
docker model run hf.co/microsoft/Phi-3-small-128k-instruct
```

flash Attention Error while inference

by hackint0sh - opened May 23, 2024

Discussion

hackint0sh

May 23, 2024

Getting the error

AssertionError: Flash Attention is not available, but is needed for dense attention
currently using the Nvidia A10G GPU

Library Installed
!pip install git+https://github.com/huggingface/transformers
!pip install tiktoken==0.6.0 triton==2.3.0

other Lib
!pip install einops accelerate bitsandbytes

zwelz

May 23, 2024

•

edited May 23, 2024

Currently running into this as well (running on 4 a100s).

Actively installing flash-attn to see if this fixes it (but can't get ninja to work for fast compile time so its slow).
https://github.com/Dao-AILab/flash-attention?tab=readme-ov-file#installation-and-features

(need cuda, gcc, etc for flash-attn)

zwelz

May 23, 2024

Currently running into this as well (running on 4 a100s).

Actively installing flash-attn to see if this fixes it (but can't get ninja to work for fast compile time so its slow).
https://github.com/Dao-AILab/flash-attention?tab=readme-ov-file#installation-and-features

(need cuda, gcc, etc for flash-attn)

Was able to get flash-attention to install with correct pytorch version.
https://pytorch.org/get-started/previous-versions/#v210

bapatra

Microsoft org May 23, 2024

Did installing flash-attention fix the issue ?

zwelz

May 23, 2024

•

edited May 23, 2024

Yes. I am able to run the Phi-3-small model now. I was also able to get ninja installed to reduce build time for flash-attention to <3 minutes (12 cores, 120GB RAM).

bapatra

Microsoft org May 23, 2024

Awesome ! Closing this out for now.
Adding the torch sdpa attention to remove the hard dependency on flash-attention is one of the things that we can subsequently follow up with if this becomes a big issue for adoption. We are glad for your interest in phi3-small, and hope you find it useful !

bapatra changed discussion status to closed May 23, 2024

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