Skip to content

Latest commit

 

History

History
147 lines (120 loc) · 6.22 KB

README.md

File metadata and controls

147 lines (120 loc) · 6.22 KB
Raw

Stable Diffusion 2.1 FineTuning with Low-Rank Adaptation of Large Language Models for PyTorch

This directory provides scripts to fine-tune Stable Diffusion Model (2.1) which is based on latent text-to-image diffusion model and is tested and maintained by Intel® Gaudi®. For more information on training and inference of deep learning models using Intel Gaudi AI accelerator, refer to developer.habana.ai. Before you get started, make sure to review the Supported Configuration.

Model Overview

This implementation is designed to fine-tune Stable Diffusion model (stabilityai/stable-diffusion-2-1-base) by Low-rank Adaptation which is considered to be very efficient. Here the pipeline is to fine-tune CLIP + Unet + token to gain better results.

More details about LoRA and its usage with diffusion can be found at Blog and examples

How to Use

Users acknowledge and understand that the models referenced by Habana are mere examples for models that can be run on Gaudi. Users bear sole liability and responsibility to follow and comply with any third party licenses pertaining to such models, and Habana Labs disclaims and will bear no any warranty or liability with respect to users' use or compliance with such third party licenses.

Setup

Please follow the instructions provided in the Gaudi Installation Guide to set up the environment including the $PYTHON environment variable. To achieve the best performance, please follow the methods outlined in the Optimizing Training Platform guide. The guides will walk you through the process of setting up your system to run the model on Gaudi.

Clone Intel Gaudi Model-References

In the docker container, clone this repository and switch to the branch that matches your Intel Gaudi software version. You can run the hl-smi utility to determine the Intel Gaudi software version.

git clone -b [Intel Gaudi software] https://github.com/HabanaAI/Model-References
cd Model-References/PyTorch/generative_models/stable-diffusion-finetuning

Install Model Requirements

  1. In the docker container, go to the model directory:
cd Model-References/PyTorch/generative_models/stable-diffusion-finetuning
  1. Install the required packages using pip:
pip install -r requirements.txt
pip install .

Training

Model Checkpoint

The fine-tuning script internally will download checkpoints from https://huggingface.co/stabilityai/stable-diffusion-2-1-base.

Users acknowledge and understand that by downloading the checkpoint referenced herein they will be required to comply with third party licenses and rights pertaining to the checkpoint, and users will be solely liable and responsible for complying with any applicable licenses. Habana Labs disclaims any warranty or liability with respect to users' use or compliance with such third party licenses.

Dataset Preparation

For the finetuning we have used synthetic dataset.

  1. In the docker container, go to the model directory:
cd Model-References/PyTorch/generative_models/stable-diffusion-finetuning
  1. Generate synthetic dataset:
python data/scripts/gen_synth_data.py

1. Fine-tuning Stable diffusion with LoRA CLI

Single Card Training Examples

Run training on 1 HPU:

export MODEL_NAME="stabilityai/stable-diffusion-2-1-base"
export INSTANCE_DIR=<path-to-instances>
export OUTPUT_DIR=<path-to-output>

lora_pti \
  --pretrained_model_name_or_path=$MODEL_NAME  \
  --instance_data_dir=$INSTANCE_DIR \
  --output_dir=$OUTPUT_DIR \
  --use_face_segmentation_condition \
  --resolution=512 \
  --train_batch_size=7 \
  --gradient_accumulation_steps=1 \
  --learning_rate_unet=5e-5 \
  --learning_rate_ti=2e-3 \
  --color_jitter \
  --lr_scheduler="linear" --lr_scheduler_lora="linear"\
  --lr_warmup_steps=0 \
  --placeholder_tokens="<s1>|<s2>" \
  --use_template="object"\
  --save_steps=50 \
  --max_train_steps_ti=500 \
  --max_train_steps_tuning=1000 \
  --perform_inversion=True \
  --clip_ti_decay \
  --weight_decay_ti=0.000 \
  --weight_decay_lora=0.001\
  --continue_inversion \
  --continue_inversion_lr=1e-3 \
  --device="hpu" \
  --lora_rank=16 \
  --use_lazy_mode=True \
  --use_fused_adamw=True \
  --print_freq=50 \
  --use_fused_clip_norm=True \
  --lora_clip_target_modules="{'CLIPSdpaAttention'}" \

Refer to reference model to see what these parameters mean.

Supported Configuration

Validated on Intel Gaudi Software Version PyTorch Version Mode
Gaudi 1.11.0 2.0.1 Training
Gaudi 2 1.18.0 2.4.0 Training

Changelog

Script Modifications

1.18.0

  • Added new paramerter in the training command to avoid the runtime error due to transformer packaage upgrade

1.15.0

  • Added support for dynamic shapes in Stable Diffusion Finetuning.

1.13.0

  • Modified training script to support diffusers version 0.21.4.

1.12.0

  • Dynamic Shapes will be enabled by default in future releases. It is currently disabled in training script.

1.11.0

  • Dynamic Shapes will be enabled by default in future releases. It is currently enabled in training script as a temporary solution.

1.10.0

  • Modified README.
  • Enabled PyTorch autocast on Gaudi.
  • Added additional logging.
  • Added support for HPU.
  • Added FusedAdamW and FusedClipNorm.
  • Added Tensorboard logging.
  • Added device trace and memory stats reporting.
  • Added print frequency change.
  • Enabled HPU Graphs execution for host optimization.