Spaces:
Runtime error
Runtime error
File size: 5,432 Bytes
44df93e 22ba7a7 d9a9ea3 44df93e d9a9ea3 44df93e 22ba7a7 44df93e d6a404f 44df93e 22ba7a7 d9a9ea3 890d938 3095ffa 7c67c9f 890d938 d9a9ea3 3095ffa d9a9ea3 22ba7a7 d9a9ea3 d6a404f 44df93e d6a404f 44df93e d6a404f 44df93e 05397a1 d9a9ea3 44df93e 05397a1 44df93e d9a9ea3 44df93e 05397a1 44df93e 05397a1 44df93e |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 |
from htbuilder import HtmlElement, div, ul, li, br, hr, a, p, img, styles, classes, fonts
from htbuilder.units import percent, px
from htbuilder.funcs import rgba, rgb
import streamlit as st
import os
import sys
import argparse
import clip
import numpy as np
from PIL import Image
from dalle.models import Dalle
from dalle.utils.utils import set_seed, clip_score
import streamlit.components.v1 as components
import torch
from IPython.display import display
import random
def link(link, text, **style):
return a(_href=link, _target="_blank", style=styles(**style))(text)
def layout(*args):
style = """
<style>
# MainMenu {visibility: hidden;}
footer {visibility: hidden;}
.stApp { bottom: 125px; }
</style>
"""
style_div = styles(
position="fixed",
left=0,
bottom=0,
margin=px(0, 0, 0, 0),
width=percent(100),
color="black",
text_align="center",
height="auto",
opacity=1
)
style_hr = styles(
display="block",
margin=px(8, 8, "auto", "auto"),
border_style="inset",
border_width=px(2)
)
body = p()
foot = div(
style=style_div
)(
hr(
style=style_hr
),
body
)
st.markdown(style, unsafe_allow_html=True)
for arg in args:
if isinstance(arg, str):
body(arg)
elif isinstance(arg, HtmlElement):
body(arg)
st.markdown(str(foot), unsafe_allow_html=True)
def footer():
myargs = [
"Created by ",
link("https://jonathanmalott.com", "Jonathan Malott"),
br(),
link("https://bridgingbarriers.utexas.edu/good-systems", "Good Systems Grand Challenge"),
", The University of Texas at Austin.",
" Advised by Dr. Junfeng Jiao.",
br(),
br(),
]
layout(*myargs)
components.html(
"""
<!-- Global site tag (gtag.js) - Google Analytics -->
<script async src="https://www.googletagmanager.com/gtag/js?id=G-SB6NJ9DQS7"></script>
<script>
window.dataLayer = window.dataLayer || [];
function gtag(){dataLayer.push(arguments);}
gtag('js', new Date());
gtag('config', 'G-SB6NJ9DQS7');
</script>
"""
)
from min_dalle import MinDalle
def generate2(prompt,crazy,k):
mm = MinDalle(
models_root='./pretrained',
dtype=torch.float32,
device='cpu',
is_mega=False,
is_reusable=False
)
image = mm.generate_image(
text=prompt,
seed=np.random.randint(0,10000),
grid_size=1,
is_seamless=False,
temperature=crazy,
top_k=k,#2128,
supercondition_factor=16,
is_verbose=False
)
item = {}
item['prompt'] = prompt
item['crazy'] = crazy
item['k'] = k
item['image'] = image
st.session_state.results.append(item)
model = False
def generate(prompt,crazy,k):
global model
device = 'cpu'
if(model == False):
model = Dalle.from_pretrained('minDALL-E/1.3B') # This will automatically download the pretrained model.
model.to(device=device)
num_candidates = 1
images = []
set_seed(np.random.randint(0,10000))
# Sampling
newPrompt = prompt
if("architecture" not in prompt.lower() ):
newPrompt += " architecture"
images = model.sampling(prompt=newPrompt,
top_k=256,
top_p=None,
softmax_temperature=crazy,
num_candidates=num_candidates,
device=device).cpu().numpy()
images = np.transpose(images, (0, 2, 3, 1))
# CLIP Re-ranking
model_clip, preprocess_clip = clip.load("ViT-B/32", device=device)
model_clip.to(device=device)
rank = clip_score(prompt=newPrompt,
images=images,
model_clip=model_clip,
preprocess_clip=preprocess_clip,
device=device)
result = images[rank]
item = {}
item['prompt'] = prompt
item['crazy'] = crazy
item['k'] = 20
item['image'] = Image.fromarray((result*255).astype(np.uint8))
st.session_state.results.append(item)
def drawGrid():
master = {}
for r in st.session_state.results[::-1]:
_txt = r['prompt']+" "+str(r['crazy'])+" "+str(r['k'])
if(_txt not in master):
master[_txt] = [r]
else:
master[_txt].append(r)
for i in st.session_state.images:
im = st.empty()
placeholder = st.empty()
with placeholder.container():
for m in master:
txt = master[m][0]['prompt']+" (temperature:"+ str(master[m][0]['crazy']) + ", top k:" + str(master[m][0]['k']) + ")"
st.subheader(txt)
col1, col2, col3 = st.columns(3)
for ix, item in enumerate(master[m]):
if ix % 3 == 0:
with col1:
st.session_state.images.append(st.image(item["image"]))
if ix % 3 == 1:
with col2:
st.session_state.images.append(st.image(item["image"]))
if ix % 3 == 2:
with col3:
st.session_state.images.append(st.image(item["image"]))
|