app.py

import gc
import math
import sys

from IPython import display
import torch
from torch import nn
from torch.nn import functional as F
from torchvision import transforms
from torchvision import utils as tv_utils
from torchvision.transforms import functional as TF
import gradio as gr
from git.repo.base import Repo
from os.path import exists as path_exists

if not (path_exists(f"v-diffusion-pytorch")):
    Repo.clone_from("https://github.com/crowsonkb/v-diffusion-pytorch", "v-diffusion-pytorch")
if not (path_exists(f"CLIP")):
    Repo.clone_from("https://github.com/openai/CLIP", "CLIP")
sys.path.append('v-diffusion-pytorch')


from huggingface_hub import hf_hub_download

from CLIP import clip
from diffusion import get_model, sampling, utils

class MakeCutouts(nn.Module):
    def __init__(self, cut_size, cutn, cut_pow=1.):
        super().__init__()
        self.cut_size = cut_size
        self.cutn = cutn
        self.cut_pow = cut_pow

    def forward(self, input):
        sideY, sideX = input.shape[2:4]
        max_size = min(sideX, sideY)
        min_size = min(sideX, sideY, self.cut_size)
        cutouts = []
        for _ in range(self.cutn):
            size = int(torch.rand([])**self.cut_pow * (max_size - min_size) + min_size)
            offsetx = torch.randint(0, sideX - size + 1, ())
            offsety = torch.randint(0, sideY - size + 1, ())
            cutout = input[:, :, offsety:offsety + size, offsetx:offsetx + size]
            cutout = F.adaptive_avg_pool2d(cutout, self.cut_size)
            cutouts.append(cutout)
        return torch.cat(cutouts)

cc12m_model = hf_hub_download(repo_id="multimodalart/crowsonkb-v-diffusion-cc12m-1-cfg", filename="cc12m_1_cfg.pth")
model = get_model('cc12m_1_cfg')()
_, side_y, side_x = model.shape
model.load_state_dict(torch.load(cc12m_model, map_location='cpu'))
model = model.half().cuda().eval().requires_grad_(False)
clip_model = clip.load(model.clip_model, jit=False, device='cpu')[0]
clip_model.eval().requires_grad_(False)
normalize = transforms.Normalize(mean=[0.48145466, 0.4578275, 0.40821073],
                                     std=[0.26862954, 0.26130258, 0.27577711])
make_cutouts = MakeCutouts(clip_model.visual.input_resolution, 16, 1.)
def run_all(prompt, steps, n_images, weight, clip_guided):
    import random
    seed = int(random.randint(0, 2147483647))
    target_embed = clip_model.encode_text(clip.tokenize(prompt)).float().cuda()
    clip_embed = target_embed.repeat([n_images, 1])
    def cfg_model_fn(x, t):
        """The CFG wrapper function."""
        n = x.shape[0]
        x_in = x.repeat([2, 1, 1, 1])
        t_in = t.repeat([2])
        clip_embed_repeat = target_embed.repeat([n, 1])
        clip_embed_in = torch.cat([torch.zeros_like(clip_embed_repeat), clip_embed_repeat])
        v_uncond, v_cond = model(x_in, t_in, clip_embed_in).chunk(2, dim=0)
        v = v_uncond + (v_cond - v_uncond) * weight
        return v
    
    def make_cond_model_fn(model, cond_fn):
        def cond_model_fn(x, t, **extra_args):
            with torch.enable_grad():
                x = x.detach().requires_grad_()
                v = model(x, t, **extra_args)
                alphas, sigmas = utils.t_to_alpha_sigma(t)
                pred = x * alphas[:, None, None, None] - v * sigmas[:, None, None, None]
                cond_grad = cond_fn(x, t, pred, **extra_args).detach()
                v = v.detach() - cond_grad * (sigmas[:, None, None, None] / alphas[:, None, None, None])
            return v
        return cond_model_fn
    def cond_fn(x, t, pred, clip_embed):
        if min(pred.shape[2:4]) < 256:
            pred = F.interpolate(pred, scale_factor=2, mode='bilinear', align_corners=False)
        clip_in = normalize(make_cutouts((pred + 1) / 2))
        image_embeds = clip_model.encode_image(clip_in).view([16, x.shape[0], -1])
        losses = spherical_dist_loss(image_embeds, clip_embed[None])
        loss = losses.mean(0).sum() * 500.
        grad = -torch.autograd.grad(loss, x)[0]
        return grad
    
    gc.collect()
    torch.cuda.empty_cache()
    torch.manual_seed(seed)
    x = torch.randn([n_images, 3, side_y, side_x], device='cuda')
    t = torch.linspace(1, 0, steps + 1, device='cuda')[:-1]
    #step_list = utils.get_spliced_ddpm_cosine_schedule(t)
    if model.min_t == 0:
        step_list = utils.get_spliced_ddpm_cosine_schedule(t)
    else:
        step_list = utils.get_ddpm_schedule(t)
    if(not clip_guided):
        outs = sampling.plms_sample(cfg_model_fn, x, step_list, {})#, callback=display_callback)
    else:
        extra_args = {'clip_embed': clip_embed}
        cond_fn_ = cond_fn
        model_fn = make_cond_model_fn(model, cond_fn_)
        outs = sampling.plms_sample(model_fn, x, step_list, extra_args)
    images_out = []
    for i, out in enumerate(outs):
        images_out.append(utils.to_pil_image(out))
    return(images_out)
    

##################### START GRADIO HERE ############################
#image = gr.outputs.Image(type="pil", label="Your result")
gallery = gr.Gallery(css={"height": "256px","width":"256px"})
iface = gr.Interface(
    fn=run_all, 
    inputs=[
    gr.inputs.Textbox(label="Prompt - try adding increments to your prompt such as 'oil on canvas', 'a painting', 'a book cover'",default="chalk pastel drawing of a dog wearing a funny hat"),
    gr.inputs.Slider(label="Steps - more steps can increase quality but will take longer to generate",default=40,maximum=80,minimum=1,step=1),
    gr.inputs.Slider(label="Number of images in parallel", default=2, maximum=4, minimum=1, step=1),
    gr.inputs.Slider(label="Weight - how closely the image should resemble the prompt", default=5, maximum=15, minimum=0, step=1),
    gr.inputs.Checkbox(label="CLIP Guided - improves coherence with prompt, makes it slower"),
    ], 
    outputs=gallery,
    title="Generate images from text with V-Diffusion CC12M CFG",
    description="<div>By typing a prompt and pressing submit you can generate images based on this prompt. <a href='https://github.com/crowsonkb/v-diffusion-pytorch' target='_blank'>V-Diffusion</a> is diffusion text-to-image model created by <a href='https://twitter.com/RiversHaveWings' target='_blank'>Katherine Crowson</a> and <a href='https://twitter.com/jd_pressman'>JDP</a>, trained on the <a href='https://github.com/google-research-datasets/conceptual-12m'>CC12M dataset</a>. CFG means it can generate images without CLIP Guidance - and fast. The UI to the model was assembled by <a style='color: rgb(99, 102, 241);font-weight:bold' href='https://twitter.com/multimodalart' target='_blank'>@multimodalart</a>, keep up with the <a style='color: rgb(99, 102, 241);' href='https://multimodal.art/news' target='_blank'>latest multimodal ai art news here</a> and consider <a style='color: rgb(99, 102, 241);' href='https://www.patreon.com/multimodalart' target='_blank'>supporting us on Patreon</a></div>",
    #article="<h4 style='font-size: 110%;margin-top:.5em'>Biases acknowledgment</h4><div>Despite how impressive being able to turn text into image is, beware to the fact that this model may output content that reinforces or exarcbates societal biases. According to the <a href='https://arxiv.org/abs/2112.10752' target='_blank'>Latent Diffusion paper</a>:<i> \"Deep learning modules tend to reproduce or exacerbate biases that are already present in the data\"</i>. The model was trained on an unfiltered version the LAION-400M dataset, which scrapped non-curated image-text-pairs from the internet (the exception being the the removal of illegal content) and is meant to be used for research purposes, such as this one. <a href='https://laion.ai/laion-400-open-dataset/' target='_blank'>You can read more on LAION's website</a></div><h4 style='font-size: 110%;margin-top:1em'>Who owns the images produced by this demo?</h4><div>Definetly not me! Probably you do. I say probably because the Copyright discussion about AI generated art is ongoing. So <a href='https://www.theverge.com/2022/2/21/22944335/us-copyright-office-reject-ai-generated-art-recent-entrance-to-paradise' target='_blank'>it may be the case that everything produced here falls automatically into the public domain</a>. But in any case it is either yours or is in the public domain.</div>"
    )
iface.launch(enable_queue=True)