Skip to content

Mikubill/naifu

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Naifu

naifu (or naifu-diffusion) is designed for training generative models with various configurations and features. The code in the main branch of this repository is under development and subject to change as new features are added.

Installation

To get started with Naifu, follow these steps to install the necessary dependencies:

# Clone the Naifu repository:
git clone --depth 1 https://github.com/mikubill/naifu

# Install the required Python packages:
cd naifu && pip install -r requirements.txt

Make sure you have a compatible version of Python installed (Python 3.9 or above).

Usage

Naifu provides a flexible and intuitive way to train models using various configurations. To train a model, use the trainer.py script and provide the desired configuration file as an argument.

python trainer.py --config config/<config_file>

# or (same as --config)
python trainer.py config/<config_file>

Replace <config_file> with one of the available configuration files listed below.

Configurations

Choose the appropriate configuration file based on training objectives and environment.

Train SDXL (Stable Diffusion XL) model

# prepare image data (to latents)
python scripts/encode_latents_xl.py -i <input_path> -o <encoded_path>

# sd_xl_base_1.0_0.9vae.safetensors
python trainer.py config/train_sdxl.yaml

# For huggingface model support
# stabilityai/stable-diffusion-xl-base-1.0
python trainer.py config/train_diffusers.yaml

# use original sgm loss module
python trainer.py config/train_sdxl_original.yaml

Train SDXL refiner (Stable Diffusion XL refiner) model

# stabilityai/stable-diffusion-xl-refiner-1.0
python trainer.py config/train_refiner.yaml

Train original Stable Diffusion 1.4 or 1.5 model

# runwayml/stable-diffusion-v1-5
# Note: will save in diffusers format
python trainer.py config/train_sd15.yaml

Train SDXL model with LyCORIS.

# Based on the work available at KohakuBlueleaf/LyCORIS
pip install lycoris_lora toml
python trainer.py config/train_lycoris.yaml

Use fairscale strategy for distributed data parallel sharded training

pip install fairscale
python trainer.py config/train_fairscale.yaml

Train SDXL model with Diffusion DPO
Paper: Diffusion Model Alignment Using Direct Preference Optimization (arxiv:2311.12908)

# dataset: yuvalkirstain/pickapic_v2
# Be careful tuning the resolution and dpo_betas!
# will save in diffusers format
python trainer.py config/train_dpo_diffusers.yaml # diffusers backend
python trainer.py config/train_dpo.yaml # sgm backend

Train Pixart-Alpha model
Paper: Fast Training of Diffusion Transformer for Photorealistic Text-to-Image Synthesis (arxiv:2310.00426)

# PixArt-alpha/PixArt-XL-2-1024-MS
python trainer.py config/train_pixart.yaml

Train SDXL-LCM model
Paper: Latent Consistency Models: Synthesizing High-Resolution Images with Few-Step Inference (arxiv:2310.04378)

python trainer.py config/train_lcm.yaml

Train StableCascade model (Sai)

# currently only stage_c (w/ or w/o text encoder)
python trainer.py config/train_cascade_stage_c.yaml

Train GPT2 model

# currently only stage_c (w/ or w/o text encoder)
python trainer.py config/train_gpt2.yaml

Train with Phi-1.5/2 model

python trainer.py config/train_phi2.yaml

Train language models (LLaMA, Qwen, Gemma etc)

# Note that prepare data in sharegpt/chatml format, or define your own dataset in data/text_dataset.py
# See example dataset for reference: function-calling-sharegpt
python trainer.py config/train_general_llm.yaml

Train language models with lora or qlora (For example, Mistral)

python trainer.py config/train_mistral_lora.yaml

Other branches

  • sgm - Uses the sgm to train SDXL models.
  • sd3 - Trainer for SD3 models - use with caution: may produce undesired result
  • hydit - Trainer for hunyuan dit models (v1.1 and v1.2)
  • main-archived - Contains the original naifu-diffusion code for training Stable Diffusion 1.x models.

For branches without documentation, please follow the installation instructions provided above.