Initial attempt

app.py

@@ -1,11 +1,83 @@
 import gradio as gr
-import torch 
-
-is_cuda = torch.cuda.is_available()
-def greet(name):
-    if is_cuda:
-        return "Hello cuda" + name + "!!"
-    else:
-        return "Hello ooops" + name + "!!"
-iface = gr.Interface(fn=greet, inputs="text", outputs="text")
-iface.launch()
+from PIL import Image
+import torch
+import numpy as np
+from os.path import exists as path_exists
+from git.repo.base import Repo
+from einops import rearrange
+import torchvision.transforms as transforms
+from torchvision.utils import make_grid
+
+if not (path_exists(f"rudalle-aspect-ratio")):
+    Repo.clone_from("https://github.com/shonenkov-AI/rudalle-aspect-ratio", "rudalle-aspect-ratio")
+
+import sys
+sys.path.append('./rudalle-aspect-ratio')
+from rudalle_aspect_ratio import RuDalleAspectRatio, get_rudalle_model
+from rudalle import get_vae, get_tokenizer
+from rudalle.pipelines import show
+#model_path_e = hf_hub_download(repo_id="multimodalart/compvis-latent-diffusion-text2img-large", filename="txt2img-f8-large.ckpt")
+
+device = 'cuda'
+dalle_surreal = get_rudalle_model('Surrealist_XL', fp16=True, device=device)
+dalle_real = get_rudalle_model('Malevich',fp16=True,device=device)
+dalle_emoji = get_rudalle_model('Emojich',fp16=True,device=device)
+vae, tokenizer = get_vae().to(device), get_tokenizer()
+def np_gallery(array, ncols=3):
+    nindex, height, width, intensity = array.shape
+    nrows = nindex//ncols
+    assert nindex == nrows*ncols
+    # want result.shape = (height*nrows, width*ncols, intensity)
+    result = (array.reshape(nrows, ncols, height, width, intensity)
+              .swapaxes(1,2)
+              .reshape(height*nrows, width*ncols, intensity))
+    return result
+
+def image_to_np(image):
+    return np.asarray(image)
+
+def run(prompt, aspect_ratio, model):
+    if(model=='Surrealism'):
+        dalle = dalle_surreal
+    elif(model=='Realism'):  
+        dalle = dalle_real
+    elif(model=='Emoji'):  
+        dalle = dalle_emoji
+    if(aspect_ratio == 'Square'):
+        aspect_ratio_value = 1
+        top_k = 512
+    elif(aspect_ratio == 'Horizontal'):
+        aspect_ratio_value = 32/9
+        top_k = 1024
+    elif(aspect_ratio == 'Vertical'):
+        aspect_ratio_value = 9/32
+        top_k = 512
+    rudalle_ar = RuDalleAspectRatio(
+        dalle=dalle, vae=vae, tokenizer=tokenizer,
+        aspect_ratio=aspect_ratio_value, bs=1, device=device
+    )
+    
+    _, result_pil_images = rudalle_ar.generate_images(prompt, top_k, 0.975, 1)
+    #np_images = map(image_to_np,result_pil_images)
+    #np_grid = np_gallery(np.array(list(np_images)),2)
+    #result_grid = Image.fromarray(np_grid)
+
+    return(result_pil_images[0])
+
+image = gr.outputs.Image(type="pil", label="Your result")
+iface = gr.Interface(fn=run, inputs=[
+    gr.inputs.Textbox(label="Prompt (if not in Russian, it will be automatically translated to Russian)",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=45,maximum=50,minimum=1,step=1),
+    gr.inputs.Radio(label="Aspect Ratio", choices=["Square", "Horizontal", "Vertical"],default="Horizontal"),
+    gr.inputs.Dropdown(label="Model", choices=["Surrealism","Realism", "Emojis"], default="Surrealism")
+    #gr.inputs.Radio(label="Height", choices=[32,64,128,256,512],default=256),
+    #gr.inputs.Slider(label="Images - How many images you wish to generate", default=2, step=1, minimum=1, maximum=4),
+    #gr.inputs.Slider(label="Diversity scale - How different from one another you wish the images to be",default=5.0, minimum=1.0, maximum=15.0),
+    #gr.inputs.Slider(label="ETA - between 0 and 1. Lower values can provide better quality, higher values can be more diverse",default=0.0,minimum=0.0, maximum=1.0,step=0.1),
+    ], 
+    outputs=image,
+    #css=css,
+    title="Generate images from text with ruDALLE",
+    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'>ruDALLE</a> is an open source text-to-image model, this Arbitrary Aspect ration implementation was created by <a href='https://github.com/shonenkov-AI' target='_blank'>Alex Shonenkov</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 models are meant to be used for research purposes, such as this one.</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)

requirements.txt

@@ -1 +1,6 @@
-torch
+torch
+torchvision
+gitpython
+rudalle==1.0.0
+einops
+numpy