Home

Awesome

Kandinsky 2.2

Open In Colab — Inference example

Open In Colab — Fine-tuning with LoRA

<p align="left"> <img src="./content/kand_22.png" width="60%"> </p>

Description:

Kandinsky 2.2 brings substantial improvements upon its predecessor, Kandinsky 2.1, by introducing a new, more powerful image encoder - CLIP-ViT-G and the ControlNet support.

The switch to CLIP-ViT-G as the image encoder significantly increases the model's capability to generate more aesthetic pictures and better understand text, thus enhancing the model's overall performance.

The addition of the ControlNet mechanism allows the model to effectively control the process of generating images. This leads to more accurate and visually appealing outputs and opens new possibilities for text-guided image manipulation.

Architecture details:

Сheckpoints:

Inference regimes

<p align="left"> <img src="./content/kand_22_setting.png" width="60%"> </p>

How to use:

Check our jupyter notebooks with examples in ./notebooks folder

1. text2image

from kandinsky2 import get_kandinsky2
model = get_kandinsky2('cuda', task_type='text2img', model_version='2.2')
images = model.generate_text2img(
    "red cat, 4k photo", 
    decoder_steps=50,
    batch_size=1, 
    h=1024,
    w=768,
)

Kandinsky 2.1

Framework: PyTorch Huggingface space Open In Colab

Habr post

Demo

pip install "git+https://github.com/ai-forever/Kandinsky-2.git"

Model architecture:

<p align="left"> <img src="./content/kandinsky21.png" width="80%"> </p>

Kandinsky 2.1 inherits best practicies from Dall-E 2 and Latent diffusion, while introducing some new ideas.

As text and image encoder it uses CLIP model and diffusion image prior (mapping) between latent spaces of CLIP modalities. This approach increases the visual performance of the model and unveils new horizons in blending images and text-guided image manipulation.

For diffusion mapping of latent spaces we use transformer with num_layers=20, num_heads=32 and hidden_size=2048.

Other architecture parts:

Kandinsky 2.1 was trained on a large-scale image-text dataset LAION HighRes and fine-tuned on our internal datasets.

How to use:

Check our jupyter notebooks with examples in ./notebooks folder

1. text2image

from kandinsky2 import get_kandinsky2
model = get_kandinsky2('cuda', task_type='text2img', model_version='2.1', use_flash_attention=False)
images = model.generate_text2img(
    "red cat, 4k photo", 
    num_steps=100,
    batch_size=1, 
    guidance_scale=4,
    h=768, w=768,
    sampler='p_sampler', 
    prior_cf_scale=4,
    prior_steps="5"
)

prompt: "Einstein in space around the logarithm scheme"

2. image fuse

from kandinsky2 import get_kandinsky2
from PIL import Image
model = get_kandinsky2('cuda', task_type='text2img', model_version='2.1', use_flash_attention=False)
images_texts = ['red cat', Image.open('img1.jpg'), Image.open('img2.jpg'), 'a man']
weights = [0.25, 0.25, 0.25, 0.25]
images = model.mix_images(
    images_texts, 
    weights, 
    num_steps=150,
    batch_size=1, 
    guidance_scale=5,
    h=768, w=768,
    sampler='p_sampler', 
    prior_cf_scale=4,
    prior_steps="5"
)

3. inpainting

from kandinsky2 import get_kandinsky2
from PIL import Image
import numpy as np

model = get_kandinsky2('cuda', task_type='inpainting', model_version='2.1', use_flash_attention=False)
init_image = Image.open('img.jpg')
mask = np.ones((768, 768), dtype=np.float32)
mask[:,:550] =  0
images = model.generate_inpainting(
    'man 4k photo', 
    init_image, 
    mask, 
    num_steps=150,
    batch_size=1, 
    guidance_scale=5,
    h=768, w=768,
    sampler='p_sampler', 
    prior_cf_scale=4,
    prior_steps="5"
)

Kandinsky 2.0

Framework: PyTorch Huggingface space Open In Colab

Habr post

Demo

pip install "git+https://github.com/ai-forever/Kandinsky-2.git"

Model architecture:

It is a latent diffusion model with two multilingual text encoders:

These encoders and multilingual training datasets unveil the real multilingual text-to-image generation experience!

Kandinsky 2.0 was trained on a large 1B multilingual set, including samples that we used to train Kandinsky.

In terms of diffusion architecture Kandinsky 2.0 implements UNet with 1.2B parameters.

Kandinsky 2.0 architecture overview:

How to use:

Check our jupyter notebooks with examples in ./notebooks folder

1. text2img

from kandinsky2 import get_kandinsky2

model = get_kandinsky2('cuda', task_type='text2img')
images = model.generate_text2img('A teddy bear на красной площади', batch_size=4, h=512, w=512, num_steps=75, denoised_type='dynamic_threshold', dynamic_threshold_v=99.5, sampler='ddim_sampler', ddim_eta=0.05, guidance_scale=10)

prompt: "A teddy bear на красной площади"

2. inpainting

from kandinsky2 import get_kandinsky2
from PIL import Image
import numpy as np

model = get_kandinsky2('cuda', task_type='inpainting')
init_image = Image.open('image.jpg')
mask = np.ones((512, 512), dtype=np.float32)
mask[100:] =  0
images = model.generate_inpainting('Девушка в красном платье', init_image, mask, num_steps=50, denoised_type='dynamic_threshold', dynamic_threshold_v=99.5, sampler='ddim_sampler', ddim_eta=0.05, guidance_scale=10)

prompt: "Девушка в красном платье"

3. img2img

from kandinsky2 import get_kandinsky2
from PIL import Image

model = get_kandinsky2('cuda', task_type='img2img')
init_image = Image.open('image.jpg')
images = model.generate_img2img('кошка', init_image, strength=0.8, num_steps=50, denoised_type='dynamic_threshold', dynamic_threshold_v=99.5, sampler='ddim_sampler', ddim_eta=0.05, guidance_scale=10)

Authors