Instructions to use 4iqq/phi-3-mini-instruct-128K-APPS-F16 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use 4iqq/phi-3-mini-instruct-128K-APPS-F16 with Transformers:

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

pipe = pipeline("text-generation", model="4iqq/phi-3-mini-instruct-128K-APPS-F16", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("4iqq/phi-3-mini-instruct-128K-APPS-F16", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("4iqq/phi-3-mini-instruct-128K-APPS-F16", 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=256)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use 4iqq/phi-3-mini-instruct-128K-APPS-F16 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "4iqq/phi-3-mini-instruct-128K-APPS-F16"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "4iqq/phi-3-mini-instruct-128K-APPS-F16",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/4iqq/phi-3-mini-instruct-128K-APPS-F16

SGLang

How to use 4iqq/phi-3-mini-instruct-128K-APPS-F16 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 "4iqq/phi-3-mini-instruct-128K-APPS-F16" \
    --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": "4iqq/phi-3-mini-instruct-128K-APPS-F16",
		"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 "4iqq/phi-3-mini-instruct-128K-APPS-F16" \
        --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": "4iqq/phi-3-mini-instruct-128K-APPS-F16",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use 4iqq/phi-3-mini-instruct-128K-APPS-F16 with Docker Model Runner:
```
docker model run hf.co/4iqq/phi-3-mini-instruct-128K-APPS-F16
```

# phi-3-mini-instruct-128K-APPS-F16

Fine-tuned Phi-3-mini-128K-instruct model specialized for reasoning and coding tasks.

## 🚀 Model Details

- **Base Model**: [microsoft/Phi-3-mini-128k-instruct](https://huggingface.co/microsoft/Phi-3-mini-128k-instruct)
- **Adapter Used**: [AdnanRiaz107/CodePhi-3-mini-128k-instruct-APPS](https://huggingface.co/AdnanRiaz107/CodePhi-3-mini-128k-instruct-APPS)
- **Architecture**: Transformer-based language model
- **Context Length**: 128K tokens
- **Specialization**: Enhanced for complex reasoning and programming tasks

## 📊 Base Model Specifications

For complete technical specifications, hardware requirements, and performance characteristics, please refer to the official base model repository:
**[microsoft/Phi-3-mini-128k-instruct](https://huggingface.co/microsoft/Phi-3-mini-128k-instruct)**

## 🛠️ Training Approach

This model was created by applying the **CodePhi-3-mini-128k-instruct-APPS** adapter to the base Phi-3 model, further optimized for coding and reasoning tasks while maintaining the original 128K context window.

## 🔧 Usage

### Direct Inference
```python
from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained(
    "4iqq/phi-3-mini-instruct-128K-APPS-F16",
    torch_dtype=torch.float16,
    device_map="auto",
    trust_remote_code=True
)
tokenizer = AutoTokenizer.from_pretrained(
    "4iqq/phi-3-mini-instruct-128K-APPS-F16",
    trust_remote_code=True
)

Convert to GGUF

python convert-hf-to-gguf.py 4iqq/phi-3-mini-instruct-128K-APPS-F16 --outtype f16

Further Fine-tuning

from peft import PeftModel, PeftConfig

model = PeftModel.from_pretrained(
    "microsoft/Phi-3-mini-128k-instruct",
    "4iqq/phi-3-mini-instruct-128K-APPS-F16"
)

📁 Repository Structure

This repository contains:

· Sharded model weights (model-0000x-of-0000x.safetensors) · Complete tokenizer files · Model configuration · Training adapters for further fine-tuning

🙏 Acknowledgments

· Microsoft for the base Phi-3-mini-128k-instruct model · AdnanRiaz107 for the original CodePhi-3 adapter

⚠️ Note

Model weights are provided in sharded format to support both:

· Direct GGUF conversion · Additional fine-tuning · Flexible deployment options

📄 License

Inherited from the base model - refer to microsoft/Phi-3-mini-128k-instruct for license details.

Downloads last month: 21

Safetensors

Model size

4B params

Tensor type

F16

Model tree for 4iqq/phi-3-mini-instruct-128K-APPS-F16

Base model

microsoft/Phi-3-mini-128k-instruct

Finetuned

(44)

this model