app.py

import gradio as gr
import torch
import requests
from io import BytesIO
from diffusers import StableDiffusionPipeline
from diffusers import DDIMScheduler
from utils import *
# from inversion_utils import *
from torch import autocast, inference_mode
import re


############################################################################################################################################################################
import torch
import os
from tqdm import tqdm
from PIL import Image, ImageDraw ,ImageFont
from matplotlib import pyplot as plt
import torchvision.transforms as T
import os
import yaml
import numpy as np
import gradio as gr


def load_512(image_path, left=0, right=0, top=0, bottom=0, device=None):
    if type(image_path) is str:
        image = np.array(Image.open(image_path).convert('RGB'))[:, :, :3]
    else:
        image = image_path
    h, w, c = image.shape
    left = min(left, w-1)
    right = min(right, w - left - 1)
    top = min(top, h - left - 1)
    bottom = min(bottom, h - top - 1)
    image = image[top:h-bottom, left:w-right]
    h, w, c = image.shape
    if h < w:
        offset = (w - h) // 2
        image = image[:, offset:offset + h]
    elif w < h:
        offset = (h - w) // 2
        image = image[offset:offset + w]
    image = np.array(Image.fromarray(image).resize((512, 512)))
    image = torch.from_numpy(image).float() / 127.5 - 1
    image = image.permute(2, 0, 1).unsqueeze(0).to(device)

    return image 


def load_real_image(folder = "data/", img_name = None, idx = 0, img_size=512, device='cuda'):
    from PIL import Image
    from glob import glob
    if img_name is not None:
        path = os.path.join(folder, img_name)
    else:
        path = glob(folder + "*")[idx]

    img = Image.open(path).resize((img_size,
                                    img_size))

    img = pil_to_tensor(img).to(device)

    if img.shape[1]== 4:
        img = img[:,:3,:,:]
    return img

def mu_tilde(model, xt,x0, timestep):
    "mu_tilde(x_t, x_0) DDPM paper eq. 7"
    prev_timestep = timestep - model.scheduler.config.num_train_timesteps // model.scheduler.num_inference_steps
    alpha_prod_t_prev = model.scheduler.alphas_cumprod[prev_timestep] if prev_timestep >= 0 else model.scheduler.final_alpha_cumprod
    alpha_t = model.scheduler.alphas[timestep]
    beta_t = 1 - alpha_t 
    alpha_bar = model.scheduler.alphas_cumprod[timestep]
    return ((alpha_prod_t_prev ** 0.5 * beta_t) / (1-alpha_bar)) * x0 +  ((alpha_t**0.5 *(1-alpha_prod_t_prev)) / (1- alpha_bar))*xt

def sample_xts_from_x0(model, x0, num_inference_steps=50):
    """
    Samples from P(x_1:T|x_0)
    """
    # torch.manual_seed(43256465436)
    alpha_bar = model.scheduler.alphas_cumprod
    sqrt_one_minus_alpha_bar = (1-alpha_bar) ** 0.5
    alphas = model.scheduler.alphas
    betas = 1 - alphas
    variance_noise_shape = (
            num_inference_steps,
            model.unet.in_channels, 
            model.unet.sample_size,
            model.unet.sample_size)
    
    timesteps = model.scheduler.timesteps.to(model.device)
    t_to_idx = {int(v):k for k,v in enumerate(timesteps)}
    xts = torch.zeros(variance_noise_shape).to(x0.device)
    for t in reversed(timesteps):
        idx = t_to_idx[int(t)]
        xts[idx] = x0 * (alpha_bar[t] ** 0.5) + torch.randn_like(x0) * sqrt_one_minus_alpha_bar[t]
    xts = torch.cat([xts, x0 ],dim = 0)

    return xts

def encode_text(model, prompts):
    text_input = model.tokenizer(
        prompts,
        padding="max_length",
        max_length=model.tokenizer.model_max_length, 
        truncation=True,
        return_tensors="pt",
    )
    with torch.no_grad():
        text_encoding = model.text_encoder(text_input.input_ids.to(model.device))[0]
    return text_encoding

def forward_step(model, model_output, timestep, sample):
    next_timestep = min(model.scheduler.config.num_train_timesteps - 2,
                        timestep + model.scheduler.config.num_train_timesteps // model.scheduler.num_inference_steps)

    # 2. compute alphas, betas
    alpha_prod_t = model.scheduler.alphas_cumprod[timestep]
    # alpha_prod_t_next = self.scheduler.alphas_cumprod[next_timestep] if next_ltimestep >= 0 else self.scheduler.final_alpha_cumprod

    beta_prod_t = 1 - alpha_prod_t

    # 3. compute predicted original sample from predicted noise also called
    # "predicted x_0" of formula (12) from https://arxiv.org/pdf/2010.02502.pdf
    pred_original_sample = (sample - beta_prod_t ** (0.5) * model_output) / alpha_prod_t ** (0.5)

    # 5. TODO: simple noising implementatiom
    next_sample = model.scheduler.add_noise(pred_original_sample,
                                    model_output,
                                    torch.LongTensor([next_timestep]))
    return next_sample


def get_variance(model, timestep): #, prev_timestep):
    prev_timestep = timestep - model.scheduler.config.num_train_timesteps // model.scheduler.num_inference_steps
    alpha_prod_t = model.scheduler.alphas_cumprod[timestep]
    alpha_prod_t_prev = model.scheduler.alphas_cumprod[prev_timestep] if prev_timestep >= 0 else model.scheduler.final_alpha_cumprod
    beta_prod_t = 1 - alpha_prod_t
    beta_prod_t_prev = 1 - alpha_prod_t_prev
    variance = (beta_prod_t_prev / beta_prod_t) * (1 - alpha_prod_t / alpha_prod_t_prev)
    return variance

def inversion_forward_process(model, x0, 
                            etas = None,    
                            prog_bar = False,
                            prompt = "",
                            cfg_scale = 3.5,
                            num_inference_steps=50, eps = None
                             ):

    if not prompt=="":
        text_embeddings = encode_text(model, prompt)
    uncond_embedding = encode_text(model, "")
    timesteps = model.scheduler.timesteps.to(model.device)
    variance_noise_shape = (
        num_inference_steps,
        model.unet.in_channels, 
        model.unet.sample_size,
        model.unet.sample_size)
    if etas is None or (type(etas) in [int, float] and etas == 0):
        eta_is_zero = True
        zs = None
    else:
        eta_is_zero = False
        if type(etas) in [int, float]: etas = [etas]*model.scheduler.num_inference_steps
        xts = sample_xts_from_x0(model, x0, num_inference_steps=num_inference_steps)
        alpha_bar = model.scheduler.alphas_cumprod
        zs = torch.zeros(size=variance_noise_shape, device=model.device)
        
    t_to_idx = {int(v):k for k,v in enumerate(timesteps)}
    xt = x0
    op = tqdm(reversed(timesteps)) if prog_bar else reversed(timesteps)

    for t in op:
        idx = t_to_idx[int(t)]
        # 1. predict noise residual
        if not eta_is_zero:
            xt = xts[idx][None]
                    
        with torch.no_grad():
            out = model.unet.forward(xt, timestep =  t, encoder_hidden_states = uncond_embedding)
            if not prompt=="":
                cond_out = model.unet.forward(xt, timestep=t, encoder_hidden_states = text_embeddings)

        if not prompt=="":
            ## classifier free guidance
            noise_pred = out.sample + cfg_scale * (cond_out.sample - out.sample)
        else:
            noise_pred = out.sample

        if eta_is_zero:
            # 2. compute more noisy image and set x_t -> x_t+1
            xt = forward_step(model, noise_pred, t, xt)

        else: 
            xtm1 =  xts[idx+1][None]
            # pred of x0
            pred_original_sample = (xt - (1-alpha_bar[t])  ** 0.5 * noise_pred ) / alpha_bar[t] ** 0.5
            
            # direction to xt
            prev_timestep = t - model.scheduler.config.num_train_timesteps // model.scheduler.num_inference_steps
            alpha_prod_t_prev = model.scheduler.alphas_cumprod[prev_timestep] if prev_timestep >= 0 else model.scheduler.final_alpha_cumprod
            
            variance = get_variance(model, t)
            pred_sample_direction = (1 - alpha_prod_t_prev - etas[idx] * variance ) ** (0.5) * noise_pred

            mu_xt = alpha_prod_t_prev ** (0.5) * pred_original_sample + pred_sample_direction
            
            z = (xtm1 - mu_xt ) / ( etas[idx] * variance ** 0.5 )
            zs[idx] = z

            # correction to avoid error accumulation
            xtm1 = mu_xt + ( etas[idx] * variance ** 0.5 )*z
            xts[idx+1] = xtm1

    if not zs is None: 
        zs[-1] = torch.zeros_like(zs[-1]) 

    return xt, zs, xts


def reverse_step(model, model_output, timestep, sample, eta = 0, variance_noise=None):
    # 1. get previous step value (=t-1)
    prev_timestep = timestep - model.scheduler.config.num_train_timesteps // model.scheduler.num_inference_steps
    # 2. compute alphas, betas
    alpha_prod_t = model.scheduler.alphas_cumprod[timestep]
    alpha_prod_t_prev = model.scheduler.alphas_cumprod[prev_timestep] if prev_timestep >= 0 else model.scheduler.final_alpha_cumprod
    beta_prod_t = 1 - alpha_prod_t
    # 3. compute predicted original sample from predicted noise also called
    # "predicted x_0" of formula (12) from https://arxiv.org/pdf/2010.02502.pdf
    pred_original_sample = (sample - beta_prod_t ** (0.5) * model_output) / alpha_prod_t ** (0.5)
    # 5. compute variance: "sigma_t(η)" -> see formula (16)
    # σ_t = sqrt((1 − α_t−1)/(1 − α_t)) * sqrt(1 − α_t/α_t−1)    
    # variance = self.scheduler._get_variance(timestep, prev_timestep)
    variance = get_variance(model, timestep) #, prev_timestep)
    std_dev_t = eta * variance ** (0.5)
    # Take care of asymetric reverse process (asyrp)
    model_output_direction = model_output
    # 6. compute "direction pointing to x_t" of formula (12) from https://arxiv.org/pdf/2010.02502.pdf
    # pred_sample_direction = (1 - alpha_prod_t_prev - std_dev_t**2) ** (0.5) * model_output_direction
    pred_sample_direction = (1 - alpha_prod_t_prev - eta * variance) ** (0.5) * model_output_direction
    # 7. compute x_t without "random noise" of formula (12) from https://arxiv.org/pdf/2010.02502.pdf
    prev_sample = alpha_prod_t_prev ** (0.5) * pred_original_sample + pred_sample_direction
    # 8. Add noice if eta > 0
    if eta > 0:
        if variance_noise is None:
            variance_noise = torch.randn(model_output.shape, device=model.device)
        sigma_z =  eta * variance ** (0.5) * variance_noise
        prev_sample = prev_sample + sigma_z

    return prev_sample

def inversion_reverse_process(model,
                    xT, 
                    etas = 0,
                    prompts = "",
                    cfg_scales = None,
                    prog_bar = False,
                    zs = None,
                    controller=None,
                    asyrp = False
                    ):

    batch_size = len(prompts)

    cfg_scales_tensor = torch.Tensor(cfg_scales).view(-1,1,1,1).to(model.device)

    text_embeddings = encode_text(model, prompts)
    uncond_embedding = encode_text(model, [""] * batch_size)

    if etas is None: etas = 0
    if type(etas) in [int, float]: etas = [etas]*model.scheduler.num_inference_steps
    assert len(etas) == model.scheduler.num_inference_steps
    timesteps = model.scheduler.timesteps.to(model.device)

    xt = xT.expand(batch_size, -1, -1, -1)
    op = tqdm(timesteps[-zs.shape[0]:]) if prog_bar else timesteps[-zs.shape[0]:] 

    t_to_idx = {int(v):k for k,v in enumerate(timesteps[-zs.shape[0]:])}

    for t in op:
        idx = t_to_idx[int(t)]        
        ## Unconditional embedding
        with torch.no_grad():
            uncond_out = model.unet.forward(xt, timestep =  t, 
                                            encoder_hidden_states = uncond_embedding)

            ## Conditional embedding  
        if prompts:  
            with torch.no_grad():
                cond_out = model.unet.forward(xt, timestep =  t, 
                                                encoder_hidden_states = text_embeddings)
            
        
        z = zs[idx] if not zs is None else None
        z = z.expand(batch_size, -1, -1, -1)
        if prompts:
            ## classifier free guidance
            noise_pred = uncond_out.sample + cfg_scales_tensor * (cond_out.sample - uncond_out.sample)
        else: 
            noise_pred = uncond_out.sample
        # 2. compute less noisy image and set x_t -> x_t-1  
        xt = reverse_step(model, noise_pred, t, xt, eta = etas[idx], variance_noise = z)

        # interm denoised img
            with autocast("cuda"), inference_mode():
                x0_dec = sd_pipe.vae.decode(1 / 0.18215 * xt).sample
                if x0_dec.dim()<4:
                    x0_dec = x0_dec[None,:,:,:]
                interm_img = image_grid(x0_dec)
                yield interm_img
            
        if controller is not None:
            xt = controller.step_callback(xt)        
    return xt, zs
############################################################################################################################################################################

def invert(x0, prompt_src="", num_diffusion_steps=100, cfg_scale_src = 3.5, eta = 1):

  #  inverts a real image according to Algorihm 1 in https://arxiv.org/pdf/2304.06140.pdf, 
  #  based on the code in https://github.com/inbarhub/DDPM_inversion
   
  #  returns wt, zs, wts:
  #  wt - inverted latent
  #  wts - intermediate inverted latents
  #  zs - noise maps

  sd_pipe.scheduler.set_timesteps(num_diffusion_steps)

  # vae encode image
  with autocast("cuda"), inference_mode():
      w0 = (sd_pipe.vae.encode(x0).latent_dist.mode() * 0.18215).float()

  # find Zs and wts - forward process
  wt, zs, wts = inversion_forward_process(sd_pipe, w0, etas=eta, prompt=prompt_src, cfg_scale=cfg_scale_src, prog_bar=True, num_inference_steps=num_diffusion_steps)
  return wt, zs, wts



def sample(wt, zs, wts, prompt_tar="", cfg_scale_tar=15, skip=36, eta = 1):

    # reverse process (via Zs and wT)
    w0, _ = inversion_reverse_process(sd_pipe, xT=wts[skip], etas=eta, prompts=[prompt_tar], cfg_scales=[cfg_scale_tar], prog_bar=False, zs=zs[skip:])
    
    # vae decode image
    with autocast("cuda"), inference_mode():
        x0_dec = sd_pipe.vae.decode(1 / 0.18215 * w0).sample
    if x0_dec.dim()<4:
        x0_dec = x0_dec[None,:,:,:]
    img = image_grid(x0_dec)
    return img

# load pipelines
# sd_model_id = "runwayml/stable-diffusion-v1-5"
# sd_model_id = "CompVis/stable-diffusion-v1-4"
sd_model_id = "stabilityai/stable-diffusion-2-base"
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
sd_pipe = StableDiffusionPipeline.from_pretrained(sd_model_id).to(device)
sd_pipe.scheduler = DDIMScheduler.from_config(sd_model_id, subfolder = "scheduler")



def get_example():
    case = [
        [
            'Examples/gnochi_mirror.jpeg', 
            '',
            'watercolor painting of a cat sitting next to a mirror',
            100,
            3.5,
            36,
            15,
            'Examples/gnochi_mirror_reconstrcution.png', 
            'Examples/gnochi_mirror_watercolor_painting.png', 
             ],]
    return case


def edit(input_image, 
            src_prompt ="", 
            tar_prompt="",
            steps=100,
            src_cfg_scale = 3.5,
            skip=36,
            seed = 0,
            left = 0,
            right = 0,
            top = 0,
            bottom = 0
):
    torch.manual_seed(seed)
     # offsets=(0,0,0,0)
    x0 = load_512(input_image, left,right, top, bottom, device)


    # invert and retrieve noise maps and latent
    wt, zs, wts = invert(x0 =x0 , prompt_src=src_prompt, num_diffusion_steps=steps, cfg_scale_src=src_cfg_scale)

    #
    xT=wts[skip]
    etas=eta
    prompts=[prompt_tar]
    cfg_scales=[cfg_scale_tar]
    prog_bar=False
    zs=zs[skip:]


    batch_size = len(prompts)

    cfg_scales_tensor = torch.Tensor(cfg_scales).view(-1,1,1,1).to(model.device)

    text_embeddings = encode_text(model, prompts)
    uncond_embedding = encode_text(model, [""] * batch_size)

    if etas is None: etas = 0
    if type(etas) in [int, float]: etas = [etas]*model.scheduler.num_inference_steps
    assert len(etas) == model.scheduler.num_inference_steps
    timesteps = model.scheduler.timesteps.to(model.device)

    xt = xT.expand(batch_size, -1, -1, -1)
    op = tqdm(timesteps[-zs.shape[0]:]) if prog_bar else timesteps[-zs.shape[0]:] 

    t_to_idx = {int(v):k for k,v in enumerate(timesteps[-zs.shape[0]:])}

    for t in op:
        idx = t_to_idx[int(t)]        
        ## Unconditional embedding
        with torch.no_grad():
            uncond_out = model.unet.forward(xt, timestep =  t, 
                                            encoder_hidden_states = uncond_embedding)

            ## Conditional embedding  
        if prompts:  
            with torch.no_grad():
                cond_out = model.unet.forward(xt, timestep =  t, 
                                                encoder_hidden_states = text_embeddings)
            
        
        z = zs[idx] if not zs is None else None
        z = z.expand(batch_size, -1, -1, -1)
        if prompts:
            ## classifier free guidance
            noise_pred = uncond_out.sample + cfg_scales_tensor * (cond_out.sample - uncond_out.sample)
        else: 
            noise_pred = uncond_out.sample
        # 2. compute less noisy image and set x_t -> x_t-1  
        xt = reverse_step(model, noise_pred, t, xt, eta = etas[idx], variance_noise = z)

        # interm denoised img
        with autocast("cuda"), inference_mode():
            x0_dec = sd_pipe.vae.decode(1 / 0.18215 * xt).sample
            if x0_dec.dim()<4:
                x0_dec = x0_dec[None,:,:,:]
                interm_img = image_grid(x0_dec)
                yield interm_img
      
    return interm_img
    
    # # vae decode image
    # with autocast("cuda"), inference_mode():
    #     x0_dec = sd_pipe.vae.decode(1 / 0.18215 * w0).sample
    # if x0_dec.dim()<4:
    #     x0_dec = x0_dec[None,:,:,:]
    # img = image_grid(x0_dec)
    # return img

    # output = sample(wt, zs, wts, prompt_tar=tar_prompt)

    # return output





########
# demo #
########
                        
intro = """
<h1 style="font-weight: 1400; text-align: center; margin-bottom: 7px;">
   Edit Friendly DDPM Inversion
</h1>
<p style="font-size: 0.9rem; text-align: center; margin: 0rem; line-height: 1.2em; margin-top:1em">
<a href="https://arxiv.org/abs/2301.12247" style="text-decoration: underline;" target="_blank">An Edit Friendly DDPM Noise Space:
Inversion and Manipulations </a> 
<p/>
<p style="font-size: 0.9rem; margin: 0rem; line-height: 1.2em; margin-top:1em">
For faster inference without waiting in queue, you may duplicate the space and upgrade to GPU in settings.
<a href="https://huggingface.co/spaces/LinoyTsaban/ddpm_sega?duplicate=true">
<img style="margin-top: 0em; margin-bottom: 0em" src="https://bit.ly/3gLdBN6" alt="Duplicate Space"></a>
<p/>"""
with gr.Blocks() as demo:
    gr.HTML(intro)
    with gr.Row():
        src_prompt = gr.Textbox(lines=1, label="Source Prompt", interactive=True, placeholder="optional: describe the original image")
        tar_prompt = gr.Textbox(lines=1, label="Target Prompt", interactive=True, placeholder="optional: describe the target image")

    with gr.Row():
        input_image = gr.Image(label="Input Image", interactive=True)
        input_image.style(height=512, width=512)
        inverted_image = gr.Image(label=f"Reconstructed Image", interactive=False)
        inverted_image.style(height=512, width=512)
        output_image = gr.Image(label=f"Edited Image", interactive=False)
        output_image.style(height=512, width=512)


    with gr.Row():
        with gr.Column(scale=1, min_width=100):
            invert_button = gr.Button("Invert")
        with gr.Column(scale=1, min_width=100):
            edit_button = gr.Button("Edit")


    with gr.Accordion("Advanced Options", open=False):
        with gr.Row():
            with gr.Column():
                #inversion
                steps = gr.Number(value=100, precision=0, label="Num Diffusion Steps", interactive=True)
                src_cfg_scale = gr.Slider(minimum=1, maximum=15, value=3.5, label=f"Source Guidance Scale", interactive=True)
      
                # reconstruction
                skip = gr.Slider(minimum=0, maximum=40, value=36, precision=0, label="Skip Steps", interactive=True)
                tar_cfg_scale = gr.Slider(minimum=7, maximum=18,value=15, label=f"Target Guidance Scale", interactive=True)
                seed = gr.Number(value=0, precision=0, label="Seed", interactive=True)

            #shift
            with gr.Column():
                left = gr.Number(value=0, precision=0, label="Left Shift", interactive=True)
                right = gr.Number(value=0, precision=0, label="Right Shift", interactive=True)
                top = gr.Number(value=0, precision=0, label="Top Shift", interactive=True)
                bottom = gr.Number(value=0, precision=0, label="Bottom Shift", interactive=True)

            
          

    # gr.Markdown(help_text)

    invert_button.click(
        fn=edit,
        inputs=[input_image, 
                    src_prompt, 
                    src_prompt,
                    steps,
                    src_cfg_scale,
                    skip,
                    seed,
                    left,
                    right,
                    top,
                    bottom
        ],
        outputs = [inverted_image],
    )

    edit_button.click(
        fn=edit,
        inputs=[input_image, 
            src_prompt, 
            tar_prompt,
            steps,
            src_cfg_scale,
            skip,
            seed,
            left,
            right,
            top,
            bottom
        ],
        outputs=[output_image],
    )


    gr.Examples(
        label='Examples', 
        examples=get_example(), 
        inputs=[input_image, src_prompt, tar_prompt, steps,
                    src_cfg_scale,
                    skip,
                    tar_cfg_scale,
                    inverted_image, output_image
               ],
        outputs=[inverted_image,output_image ],
        # fn=edit,
        # cache_examples=True
    )



demo.queue()
demo.launch(share=False)