Shivenys

phi3-mini-lora-tutorial

README

License: apache-2.0

Base Model

Base Model: microsoft/Phi-3-mini-4k-instruct
Fine-Tuning Method: QLoRA (4-bit Quantization + LoRA)
Framework: Hugging Face Transformers
PEFT Library: PEFT
Trainer: TRL SFTTrainer

Model Details

Table with columns: Parameter, Value
Parameter	Value
Base Model	Phi-3 Mini 4K Instruct
Fine-Tuning Method	QLoRA
LoRA Rank (r)	8
LoRA Alpha	16
LoRA Dropout	0.05
Task	Causal Language Modeling
Quantization	4-bit NF4
Compute Type	FP16

Training Configuration

Epochs: 5
Batch Size: 1
Learning Rate: 2e-4
Max Sequence Length: 256
Optimizer: AdamW (default TRL optimizer)

Training Dataset

The adapter was fine-tuned on a small custom instruction-response dataset containing sample examples such as:

Who created Python?
What is AI?
What is LoRA?
What is the capital of France?

This dataset was intentionally kept small for educational purposes to demonstrate the end-to-end fine-tuning process.

Usage

Install the required libraries:

bash
pip install transformers peft bitsandbytes accelerate torch

Load the adapter:

python
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig
from peft import PeftModel

BASE_MODEL = "microsoft/Phi-3-mini-4k-instruct"

bnb_config = BitsAndBytesConfig(
    load_in_4bit=True,
    bnb_4bit_quant_type="nf4",
    bnb_4bit_compute_dtype=torch.float16,
    bnb_4bit_use_double_quant=True,
)

tokenizer = AutoTokenizer.from_pretrained(BASE_MODEL)

base_model = AutoModelForCausalLM.from_pretrained(
    BASE_MODEL,
    quantization_config=bnb_config,
    device_map="auto",
)

model = PeftModel.from_pretrained(
    base_model,
    "YOUR_USERNAME/YOUR_REPOSITORY_NAME"
)

Generate text:

python
prompt = """### Instruction:
Who created Python?

### Response:
"""

inputs = tokenizer(prompt, return_tensors="pt").to(model.device)

outputs = model.generate(
    **inputs,
    max_new_tokens=100,
)

print(tokenizer.decode(outputs[0], skip_special_tokens=True))

Repository Structure

markdown
adapter_config.json
adapter_model.safetensors
README.md

Limitations

This adapter was trained on a very small dataset.
It is intended for educational and demonstration purposes.
It should not be considered a production-ready fine-tuned model.

Acknowledgements

Microsoft for the Phi-3 Mini 4K Instruct model
Hugging Face Transformers
Hugging Face TRL
PEFT
BitsAndBytes

License

This repository contains only the LoRA adapter.

Please follow the license terms of the original Phi-3 Mini 4K Instruct model when using the base model.

Available on FriendliAI

Dedicated Endpoints

Run this model inference on single tenant GPU with unmatched speed and reliability at scale.

Learn more

Container

Run this model inference with full control and performance in your environment.

Learn more

Model Details

Model Provider

Shivenys

Model Tree

Base

microsoft/Phi-3-mini-4k-instruct

Adapter

this model

Input Modalities

Text

Output Modalities

Text

Supported Functionality

Dedicated Endpoints

Container

Explore FriendliAI today

Get started Talk to an engineer

README

License: apache-2.0

Base Model

Base Model: microsoft/Phi-3-mini-4k-instruct
Fine-Tuning Method: QLoRA (4-bit Quantization + LoRA)
Framework: Hugging Face Transformers
PEFT Library: PEFT
Trainer: TRL SFTTrainer

Model Details

Table with columns: Parameter, Value
Parameter	Value
Base Model	Phi-3 Mini 4K Instruct
Fine-Tuning Method	QLoRA
LoRA Rank (r)	8
LoRA Alpha	16
LoRA Dropout	0.05
Task	Causal Language Modeling
Quantization	4-bit NF4
Compute Type	FP16

Training Configuration

Epochs: 5
Batch Size: 1
Learning Rate: 2e-4
Max Sequence Length: 256
Optimizer: AdamW (default TRL optimizer)

Training Dataset

The adapter was fine-tuned on a small custom instruction-response dataset containing sample examples such as:

Who created Python?
What is AI?
What is LoRA?
What is the capital of France?

This dataset was intentionally kept small for educational purposes to demonstrate the end-to-end fine-tuning process.

Usage

Install the required libraries:

bash
pip install transformers peft bitsandbytes accelerate torch

Load the adapter:

python
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig
from peft import PeftModel

BASE_MODEL = "microsoft/Phi-3-mini-4k-instruct"

bnb_config = BitsAndBytesConfig(
    load_in_4bit=True,
    bnb_4bit_quant_type="nf4",
    bnb_4bit_compute_dtype=torch.float16,
    bnb_4bit_use_double_quant=True,
)

tokenizer = AutoTokenizer.from_pretrained(BASE_MODEL)

base_model = AutoModelForCausalLM.from_pretrained(
    BASE_MODEL,
    quantization_config=bnb_config,
    device_map="auto",
)

model = PeftModel.from_pretrained(
    base_model,
    "YOUR_USERNAME/YOUR_REPOSITORY_NAME"
)

Generate text:

python
prompt = """### Instruction:
Who created Python?

### Response:
"""

inputs = tokenizer(prompt, return_tensors="pt").to(model.device)

outputs = model.generate(
    **inputs,
    max_new_tokens=100,
)

print(tokenizer.decode(outputs[0], skip_special_tokens=True))

Repository Structure

markdown
adapter_config.json
adapter_model.safetensors
README.md

Limitations

This adapter was trained on a very small dataset.
It is intended for educational and demonstration purposes.
It should not be considered a production-ready fine-tuned model.

Acknowledgements

Microsoft for the Phi-3 Mini 4K Instruct model
Hugging Face Transformers
Hugging Face TRL
PEFT
BitsAndBytes

License

This repository contains only the LoRA adapter.

Please follow the license terms of the original Phi-3 Mini 4K Instruct model when using the base model.