Instructions to use nsbendre25/Phi-3-mini-128k-instruct-ov-fp16-int4-asym with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use nsbendre25/Phi-3-mini-128k-instruct-ov-fp16-int4-asym with Transformers:

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

pipe = pipeline("text-generation", model="nsbendre25/Phi-3-mini-128k-instruct-ov-fp16-int4-asym", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("nsbendre25/Phi-3-mini-128k-instruct-ov-fp16-int4-asym", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("nsbendre25/Phi-3-mini-128k-instruct-ov-fp16-int4-asym", trust_remote_code=True, 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 nsbendre25/Phi-3-mini-128k-instruct-ov-fp16-int4-asym with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "nsbendre25/Phi-3-mini-128k-instruct-ov-fp16-int4-asym"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "nsbendre25/Phi-3-mini-128k-instruct-ov-fp16-int4-asym",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/nsbendre25/Phi-3-mini-128k-instruct-ov-fp16-int4-asym

SGLang

How to use nsbendre25/Phi-3-mini-128k-instruct-ov-fp16-int4-asym 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 "nsbendre25/Phi-3-mini-128k-instruct-ov-fp16-int4-asym" \
    --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": "nsbendre25/Phi-3-mini-128k-instruct-ov-fp16-int4-asym",
		"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 "nsbendre25/Phi-3-mini-128k-instruct-ov-fp16-int4-asym" \
        --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": "nsbendre25/Phi-3-mini-128k-instruct-ov-fp16-int4-asym",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use nsbendre25/Phi-3-mini-128k-instruct-ov-fp16-int4-asym with Docker Model Runner:
```
docker model run hf.co/nsbendre25/Phi-3-mini-128k-instruct-ov-fp16-int4-asym
```

Phi-3-128K-Instruct-ov-fp16-int4-asym

Model Description

This is a version of the original Phi-3-128K-Instruct model, converted to OpenVINO™ IR (Intermediate Representation) format for optimized inference on Intel® hardware. This model is created using the procedures detailed in the OpenVINO™ Notebooks repository.

Intended Use

This model is designed for advanced natural language understanding and generation tasks, ideal for developers and researchers in both academic and commercial settings who require efficient AI capabilities for devices with limited computational power. It is not intended for use in creating or promoting harmful or illegal content, in accordance with the guidelines outlined in the Phi-3 Acceptable Use Policy.

Licensing and Redistribution

This model is released under the MIT license.

Weight Compression Parameters

For more information on the parameters, refer to the OpenVINO™ 2024.1.0 documentation

mode: INT4_ASYM
group_size: 128
ratio: 0.8

Running Model Inference

Install packages required for using Optimum Intel integration with the OpenVINO™ backend:

pip install --upgrade --upgrade-strategy eager "optimum[openvino]"

from optimum.intel.openvino import OVModelForCausalLM
from transformers import AutoTokenizer

model_id = "microsoft/Phi-3-128K-Instruct-ov-fp32-int4-asym"

# Initialize the tokenizer and model
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = OVModelForCausalLM.from_pretrained(model_id)

pipeline = transformers.pipeline("text-generation", model=model, model_kwargs={"torch_dtype": torch.bfloat16}, device_map="auto")
pipeline("i am in paris, plan me a 2 week trip")

Downloads last month: 8