Spaces:
Runtime error
Runtime error
| import gc | |
| import math | |
| import sys | |
| from IPython import display | |
| import torch | |
| 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 | |
| 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] | |
| def run_all(prompt, steps, n_images, weight): | |
| import random | |
| seed = int(random.randint(0, 2147483647)) | |
| target_embed = clip_model.encode_text(clip.tokenize(prompt)).float().cuda() | |
| 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 | |
| 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) | |
| outs = sampling.plms_sample(cfg_model_fn, x, step_list, {})#, callback=display_callback) | |
| 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=50,maximum=250,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", default=5, maximum=15, minimum=0, step=1), | |
| #gr.inputs.Checkbox(label="CLIP Guided"), | |
| #gr.inputs.Dropdown(label="Flavor",choices=["ginger", "cumin", "holywater", "zynth", "wyvern", "aaron", "moth", "juu", "custom"]), | |
| #markdown, | |
| #gr.inputs.Dropdown(label="Style",choices=["Default","Balanced","Detailed","Consistent Creativity","Realistic","Smooth","Subtle MSE","Hyper Fast Results"],default="Hyper Fast Results"), | |
| #gr.inputs.Radio(label="Width", choices=[32,64,128,256,512],default=512), | |
| #gr.inputs.Radio(label="Height", choices=[32,64,128,256,512],default=512), | |
| ], | |
| outputs=gallery, | |
| title="Generate images from text with V-Diffusion CC12M", | |
| #description="<div>By typing a prompt and pressing submit you can generate images based on this prompt. <a href='https://github.com/CompVis/latent-diffusion' target='_blank'>Latent Diffusion</a> is a text-to-image model created by <a href='https://github.com/CompVis' target='_blank'>CompVis</a>, trained on the <a href='https://laion.ai/laion-400-open-dataset/'>LAION-400M dataset.</a><br>This UI to the model was assembled by <a style='color: rgb(245, 158, 11);font-weight:bold' href='https://twitter.com/multimodalart' target='_blank'>@multimodalart</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() |