src/ptp_sdxl/null_text_sdxl.py

from typing import Optional, Union, Tuple, List, Callable, Dict
from tqdm import tqdm
import torch
from diffusers import DDIMScheduler
from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl import (
    StableDiffusionXLPipelineOutput,
)
import torch.nn.functional as nnf
import numpy as np
import utils.ptp_nulltext.ptp_sd.ptp_utils as ptp_utils
from torch.optim.adam import Adam
from PIL import Image

device = torch.device('cuda:0') if torch.cuda.is_available() else torch.device('cpu')

NUM_DDIM_STEPS=50
GUIDANCE_SCALE=5.0

class NullInversion_sdxl:
    
    def prev_step(self, model_output: Union[torch.FloatTensor, np.ndarray], timestep: int, sample: Union[torch.FloatTensor, np.ndarray]):
        prev_timestep = timestep - self.scheduler.config.num_train_timesteps // self.scheduler.num_inference_steps
        alpha_prod_t = self.scheduler.alphas_cumprod[timestep]
        alpha_prod_t_prev = self.scheduler.alphas_cumprod[prev_timestep] if prev_timestep >= 0 else self.scheduler.final_alpha_cumprod
        beta_prod_t = 1 - alpha_prod_t
        pred_original_sample = (sample - beta_prod_t ** 0.5 * model_output) / alpha_prod_t ** 0.5
        pred_sample_direction = (1 - alpha_prod_t_prev) ** 0.5 * model_output
        prev_sample = alpha_prod_t_prev ** 0.5 * pred_original_sample + pred_sample_direction
        return prev_sample
    
    def next_step(self, model_output: Union[torch.FloatTensor, np.ndarray], timestep: int, sample: Union[torch.FloatTensor, np.ndarray]):
        timestep, next_timestep = min(timestep - self.scheduler.config.num_train_timesteps // self.scheduler.num_inference_steps, 999), timestep
        alpha_prod_t = self.scheduler.alphas_cumprod[timestep] if timestep >= 0 else self.scheduler.final_alpha_cumprod
        alpha_prod_t_next = self.scheduler.alphas_cumprod[next_timestep]
        beta_prod_t = 1 - alpha_prod_t
        next_original_sample = (sample - beta_prod_t ** 0.5 * model_output) / alpha_prod_t ** 0.5
        next_sample_direction = (1 - alpha_prod_t_next) ** 0.5 * model_output
        next_sample = alpha_prod_t_next ** 0.5 * next_original_sample + next_sample_direction
        return next_sample
    
    def get_noise_pred_single(self, latents, t, context):
        noise_pred = self.model.unet(latents, t, encoder_hidden_states=context)["sample"]
        return noise_pred

    def get_noise_pred(self, latents, t, is_forward=True, context=None):
        latents_input = torch.cat([latents] * 2)
        if context is None:
            context = self.context
        guidance_scale = 1 if is_forward else GUIDANCE_SCALE
        noise_pred = self.model.unet(latents_input, t, encoder_hidden_states=context)["sample"]
        noise_pred_uncond, noise_prediction_text = noise_pred.chunk(2)
        noise_pred = noise_pred_uncond + guidance_scale * (noise_prediction_text - noise_pred_uncond)
        if is_forward:
            latents = self.next_step(noise_pred, t, latents)
        else:
            latents = self.prev_step(noise_pred, t, latents)
        return latents

    @torch.no_grad()
    def image2latent(self, image):
        return self.model(
            prompt="",
            image=image,
            strength=0.0,
        )[0][0]

    @torch.no_grad()
    def init_prompt(self, prompt: str):
        uncond_input = self.model.tokenizer(
            [""], padding="max_length", max_length=self.model.tokenizer.model_max_length,
            return_tensors="pt"
        )
        uncond_embeddings = self.model.text_encoder(uncond_input.input_ids.to(self.model.device))[0]
        text_input = self.model.tokenizer(
            [prompt],
            padding="max_length",
            max_length=self.model.tokenizer.model_max_length,
            truncation=True,
            return_tensors="pt",
        )
        text_embeddings = self.model.text_encoder(text_input.input_ids.to(self.model.device))[0]
        self.context = torch.cat([uncond_embeddings, text_embeddings])
        self.prompt = prompt

    @torch.no_grad()
    def ddim_loop(self, latent):
        uncond_embeddings, cond_embeddings = self.context.chunk(2)
        all_latent = [latent]
        latent = latent.clone().detach()
        for i in range(NUM_DDIM_STEPS):
            t = self.model.scheduler.timesteps[len(self.model.scheduler.timesteps) - i - 1]
            noise_pred = self.get_noise_pred_single(latent, t, cond_embeddings)
            latent = self.next_step(noise_pred, t, latent)
            all_latent.append(latent)
        return all_latent

    @property
    def scheduler(self):
        return self.model.scheduler

    @torch.no_grad()
    def ddim_inversion(self, image):
        latent = self.image2latent(image)
        ddim_latents = self.ddim_loop(latent)
        return ddim_latents

    def null_optimization(self, latents, num_inner_steps, epsilon):
        uncond_embeddings, cond_embeddings = self.context.chunk(2)
        uncond_embeddings_list = []
        latent_cur = latents[-1]
        bar = tqdm(total=num_inner_steps * NUM_DDIM_STEPS)
        for i in range(NUM_DDIM_STEPS):
            uncond_embeddings = uncond_embeddings.clone().detach()
            uncond_embeddings.requires_grad = True
            optimizer = Adam([uncond_embeddings], lr=1e-2 * (1. - i / 100.))
            latent_prev = latents[len(latents) - i - 2]
            t = self.model.scheduler.timesteps[i]
            with torch.no_grad():
                noise_pred_cond = self.get_noise_pred_single(latent_cur, t, cond_embeddings)
            for j in range(num_inner_steps):
                noise_pred_uncond = self.get_noise_pred_single(latent_cur, t, uncond_embeddings)
                noise_pred = noise_pred_uncond + GUIDANCE_SCALE * (noise_pred_cond - noise_pred_uncond)
                latents_prev_rec = self.prev_step(noise_pred, t, latent_cur)
                loss = nnf.mse_loss(latents_prev_rec, latent_prev)
                optimizer.zero_grad()
                loss.backward()
                optimizer.step()
                loss_item = loss.item()
                bar.update()
                if loss_item < epsilon + i * 2e-5:
                    break
            for j in range(j + 1, num_inner_steps):
                bar.update()
            uncond_embeddings_list.append(uncond_embeddings[:1].detach())
            with torch.no_grad():
                context = torch.cat([uncond_embeddings, cond_embeddings])
                latent_cur = self.get_noise_pred(latent_cur, t, False, context)
        bar.close()
        return uncond_embeddings_list
    
    def invert(self, image: str, prompt: str, offsets=(0,0,0,0), num_inner_steps=10, early_stop_epsilon=1e-5, verbose=False):
        self.init_prompt(prompt)
        ptp_utils.register_attention_control(self.model, None)
        if verbose:
            print("DDIM inversion...")
        ddim_latents = self.ddim_inversion(image)
        if verbose:
            print("Null-text optimization...")
        uncond_embeddings = self.null_optimization(ddim_latents, num_inner_steps, early_stop_epsilon)
        output = StableDiffusionXLPipelineOutput(images=ddim_latents[-1])
        output.uncond_embeddings = uncond_embeddings
        return output
        
    
    def __init__(self, model):
        # scheduler = DDIMScheduler(beta_start=0.00085, beta_end=0.012, beta_schedule="scaled_linear", clip_sample=False,
        #                           set_alpha_to_one=False)
        self.model = model
        self.tokenizer = self.model.tokenizer
        self.model.scheduler.set_timesteps(NUM_DDIM_STEPS)
        self.prompt = None
        self.context = None