deltatuner 1.0b202310230954

Intel extension for peft with PyTorch and DENAS

Deltatuner

Deltatuner is an extension for Peft to improve LLM fine-tuning speed through multiple optimizations, including: leverage the compact model constructor DE-NAS to construct/modify the compact delta layers in a hardware-aware and train-free approach, and adding more new deltatuning algorithms.

Introduction

Key Components

• Supported parameter efficient finetuning algorithms
  • LoRA algorithm
  • Scaling and Shifting(SSF) algorithm: Scale and Shift the deep features in a pre-trained model with much less parameters to catch up with the performance of full finetuning
  • WIP on adding more algos (AdaLora etc.)
• De-Nas: Automatically construct compact and optimal delta layers with train-free and hardware-aware mode (more details here)
  • step1: Generate search space for delta layers
  • step2: Search algorithm populates delta layers for LM
  • step3: Train-free score evaluates LM with adaptive delta layers

Features

• Easy-to-use: provide package install, just need to inject few codes into the original code
• Auto-tuning: automatically select best algorithms and delta structure for finetuning model

Values

• Saving computation power: reduce the computation power and time required to fine-tune a model by reducing parameter size as well as memory footprint.
• Improve accuracy: ensure same or no accuracy regression.

Get Started

Installation

• install the python package

pip install deltatuner

Fast Fine-tuning on Base models

Below is an example of optimizing MPT model by adding the following few-lines to use the delatuner optimizations. It use the DE-NAS in delatuner to optimize a LLM with LoRA layers to a LLM with compact LoRA layers, so as to efficiently improve the LLM fine-tuning process in peak memory reduction and time speedup.

from delta import deltatuner, deltatuner_args
from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import LoraConfig, get_peft_model
from deltatuner import deltatuner, deltatuner_args

# import model from huggingface
model_id =  "mosaicml/mpt-7b"
model = AutoModelForCausalLM.from_pretrained(model_id)
tokenizer = AutoTokenizer.from_pretrained(model_id)
# adding the lora componenents with peft
config = LoraConfig()
lora_model = get_peft_model(model, config) 
# delatuner optimize the model with best lora layer configuration
deltatuning_args = deltatuner_args.DeltaTunerArguments()
deltatuner_model = deltatuner.optimize(model=lora_model, tokenizer=tokenizer, deltatuning_args=deltatuning_args)
...

API reference

In above examples, deltatuner.optimize is a python function to using deltatuner supported optimization algorithms to optimize the model.

def optimize(model, tokenizer, algo: str="auto", deltatuning_args: DeltaTunerArguments=None) -> DeltaTunerModel:
    '''
    Parameters:
        model  - a PreTrainedModel or LoraModel. Specifies what model should be optimized
        tokenizer - a tokenizer for preprocess text
        algo (str, optional) – the algorithm. Specifies what type of adapter algorithm (default: “auto”)
            "auto" – If the input model is mpt, the algorithm is ssf; elif the algorithm is lora
            "lora" – use the lora algotihm
            "ssf" – use the ssf algotithm
        deltatuning_args (optional) – the deltatuner configuration. Specifically, deltatuning_args.denas is to use the denas in the optimization (default: True)
    Return 
        DeltaTunerModel - a wrapper of model, which composed of the original properties/function together with adavance properties/function provided by deltatuner
    '''

Detailed examples

Please refer to example page for more use cases on fine-tuning other LLMs with the help of DeltaTuner.

Model supported matrix

Causal Language Modeling

Model	LoRA	SSF
GPT-2	✅
GPT-J	✅	✅
Bloom	✅	✅
OPT	✅	✅
GPT-Neo	✅	✅
Falcon	✅	✅
LLaMA	✅	✅
MPT	✅	✅