File size: 10,887 Bytes
a1b8e4d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
#@title Prepare the Concepts Library to be used

import requests
import os
import gradio as gr
import wget
import torch
from torch import autocast
from diffusers import StableDiffusionPipeline
from huggingface_hub import HfApi
from transformers import CLIPTextModel, CLIPTokenizer
from tqdm.notebook import tqdm

api = HfApi()
models_list = api.list_models(author="sd-concepts-library")
models = []

pipe = StableDiffusionPipeline.from_pretrained("CompVis/stable-diffusion-v1-4", use_auth_token=True, revision="fp16", torch_dtype=torch.float16).to("cuda")

def load_learned_embed_in_clip(learned_embeds_path, text_encoder, tokenizer, token=None):
  loaded_learned_embeds = torch.load(learned_embeds_path, map_location="cpu")
  
  # separate token and the embeds
  trained_token = list(loaded_learned_embeds.keys())[0]
  embeds = loaded_learned_embeds[trained_token]

  # cast to dtype of text_encoder
  dtype = text_encoder.get_input_embeddings().weight.dtype
  embeds.to(dtype)

  # add the token in tokenizer
  token = token if token is not None else trained_token
  num_added_tokens = tokenizer.add_tokens(token)
  i = 1
  while(num_added_tokens == 0):
    print(f"The tokenizer already contains the token {token}.")
    token = f"{token[:-1]}-{i}>"
    print(f"Attempting to add the token {token}.")
    num_added_tokens = tokenizer.add_tokens(token)
    i+=1
  
  # resize the token embeddings
  text_encoder.resize_token_embeddings(len(tokenizer))
  
  # get the id for the token and assign the embeds
  token_id = tokenizer.convert_tokens_to_ids(token)
  text_encoder.get_input_embeddings().weight.data[token_id] = embeds
  return token

print("Setting up the public library")
for model in tqdm(models_list):
  model_content = {}
  model_id = model.modelId
  model_content["id"] = model_id
  embeds_url = f"https://huggingface.co/{model_id}/resolve/main/learned_embeds.bin"
  os.makedirs(model_id,exist_ok = True)
  if not os.path.exists(f"{model_id}/learned_embeds.bin"):
    try:
      wget.download(embeds_url, out=model_id)
    except:
      continue
  token_identifier = f"https://huggingface.co/{model_id}/raw/main/token_identifier.txt"
  response = requests.get(token_identifier)
  token_name = response.text
  
  concept_type = f"https://huggingface.co/{model_id}/raw/main/type_of_concept.txt"
  response = requests.get(concept_type)
  concept_name = response.text
  model_content["concept_type"] = concept_name
  images = []
  for i in range(4):
    url = f"https://huggingface.co/{model_id}/resolve/main/concept_images/{i}.jpeg"
    image_download = requests.get(url)
    url_code = image_download.status_code
    if(url_code == 200):
      file = open(f"{model_id}/{i}.jpeg", "wb") ## Creates the file for image
      file.write(image_download.content) ## Saves file content
      file.close()
      images.append(f"{model_id}/{i}.jpeg")
  model_content["images"] = images

  learned_token = load_learned_embed_in_clip(f"{model_id}/learned_embeds.bin", pipe.text_encoder, pipe.tokenizer, token_name)
  model_content["token"] = learned_token
  models.append(model_content)
  
#@title Run the app to navigate around [the Library](https://huggingface.co/sd-concepts-library)
#@markdown Click the `Running on public URL:` result to run the Gradio app

SELECT_LABEL = "Select concept"

def title_block(title, id):
  return gr.Markdown(f"### [`{title}`](https://huggingface.co/{id})")

def image_block(image_list, concept_type):
  return gr.Gallery(
          label=concept_type, value=image_list, elem_id="gallery"
          ).style(grid=[2], height="auto")

def checkbox_block():
  checkbox = gr.Checkbox(label=SELECT_LABEL).style(container=False)
  return checkbox

def infer(text):
  with autocast("cuda"):
        images_list = pipe(
            [text]*2,
            num_inference_steps=50,
            guidance_scale=7.5
  )
  output_images = []
  for i, image in enumerate(images_list["sample"]):
    output_images.append(image)
  return output_images
  
css = '''
.gradio-container {font-family: 'IBM Plex Sans', sans-serif}
#top_title{margin-bottom: .5em}
#top_title h2{margin-bottom: 0; text-align: center}
#main_row{flex-wrap: wrap; gap: 1em; max-height: calc(100vh - 16em); overflow-y: scroll; flex-direction: row}
@media (min-width: 768px){#main_row > div{flex: 1 1 32%; margin-left: 0 !important}}
.gr-prose code::before, .gr-prose code::after {content: "" !important}
::-webkit-scrollbar {width: 10px}
::-webkit-scrollbar-track {background: #f1f1f1}
::-webkit-scrollbar-thumb {background: #888}
::-webkit-scrollbar-thumb:hover {background: #555}
.gr-button {white-space: nowrap}
.gr-button:focus {
  border-color: rgb(147 197 253 / var(--tw-border-opacity));
  outline: none;
  box-shadow: var(--tw-ring-offset-shadow), var(--tw-ring-shadow), var(--tw-shadow, 0 0 #0000);
  --tw-border-opacity: 1;
  --tw-ring-offset-shadow: var(--tw-ring-inset) 0 0 0 var(--tw-ring-offset-width) var(--tw-ring-offset-color);
  --tw-ring-shadow: var(--tw-ring-inset) 0 0 0 calc(3px var(--tw-ring-offset-width)) var(--tw-ring-color);
  --tw-ring-color: rgb(191 219 254 / var(--tw-ring-opacity));
  --tw-ring-opacity: .5;
}
#prompt_input{flex: 1 3 auto}
#prompt_area{margin-bottom: .75em}
#prompt_area > div:first-child{flex: 1 3 auto}
'''
examples = ["a <cat-toy> in <madhubani-art> style", "a mecha robot in <line-art> style", "a piano being played by <bonzi>"]
with gr.Blocks(css=css) as demo:
  state = gr.Variable({
        'selected': -1
  })
  state = {}
  def update_state(i):
        global checkbox_states
        if(checkbox_states[i]):
          checkbox_states[i] = False
          state[i] = False
        else:
          state[i] = True
          checkbox_states[i] = True
  gr.HTML('''
  <div style="text-align: center; max-width: 720px; margin: 0 auto;">
              <div
                style="
                  display: inline-flex;
                  align-items: center;
                  gap: 0.8rem;
                  font-size: 1.75rem;
                "
              >
                <svg
                  width="0.65em"
                  height="0.65em"
                  viewBox="0 0 115 115"
                  fill="none"
                  xmlns="http://www.w3.org/2000/svg"
                >
                  <rect width="23" height="23" fill="white"></rect>
                  <rect y="69" width="23" height="23" fill="white"></rect>
                  <rect x="23" width="23" height="23" fill="#AEAEAE"></rect>
                  <rect x="23" y="69" width="23" height="23" fill="#AEAEAE"></rect>
                  <rect x="46" width="23" height="23" fill="white"></rect>
                  <rect x="46" y="69" width="23" height="23" fill="white"></rect>
                  <rect x="69" width="23" height="23" fill="black"></rect>
                  <rect x="69" y="69" width="23" height="23" fill="black"></rect>
                  <rect x="92" width="23" height="23" fill="#D9D9D9"></rect>
                  <rect x="92" y="69" width="23" height="23" fill="#AEAEAE"></rect>
                  <rect x="115" y="46" width="23" height="23" fill="white"></rect>
                  <rect x="115" y="115" width="23" height="23" fill="white"></rect>
                  <rect x="115" y="69" width="23" height="23" fill="#D9D9D9"></rect>
                  <rect x="92" y="46" width="23" height="23" fill="#AEAEAE"></rect>
                  <rect x="92" y="115" width="23" height="23" fill="#AEAEAE"></rect>
                  <rect x="92" y="69" width="23" height="23" fill="white"></rect>
                  <rect x="69" y="46" width="23" height="23" fill="white"></rect>
                  <rect x="69" y="115" width="23" height="23" fill="white"></rect>
                  <rect x="69" y="69" width="23" height="23" fill="#D9D9D9"></rect>
                  <rect x="46" y="46" width="23" height="23" fill="black"></rect>
                  <rect x="46" y="115" width="23" height="23" fill="black"></rect>
                  <rect x="46" y="69" width="23" height="23" fill="black"></rect>
                  <rect x="23" y="46" width="23" height="23" fill="#D9D9D9"></rect>
                  <rect x="23" y="115" width="23" height="23" fill="#AEAEAE"></rect>
                  <rect x="23" y="69" width="23" height="23" fill="black"></rect>
                </svg>
                <h1 style="font-weight: 900; margin-bottom: 7px;">
                  Stable Diffusion Conceptualizer
                </h1>
              </div>
              <p style="margin-bottom: 10px; font-size: 94%">
                Navigate through community created concepts and styles via Stable Diffusion Textual Inversion and pick yours for inference.
                To train your own concepts and contribute to the library <a style="text-decoration: underline" href="https://colab.research.google.com/github/huggingface/notebooks/blob/main/diffusers/sd_textual_inversion_training.ipynb">check out this notebook</a>.
              </p>
            </div>
  ''')
  with gr.Row():
        with gr.Column():
          gr.Markdown('''
          ### Textual-Inversion trained [concepts library](https://huggingface.co/sd-concepts-library) navigator
          ''')
          with gr.Row(elem_id="main_row"):
                  image_blocks = []
                  for i, model in enumerate(models):
                    with gr.Box().style(border=None):
                      title_block(model["token"], model["id"])
                      image_blocks.append(image_block(model["images"], model["concept_type"]))
        with gr.Box():
                with gr.Row(elem_id="prompt_area").style(mobile_collapse=False, equal_height=True):
                    text = gr.Textbox(
                        label="Enter your prompt", placeholder="Enter your prompt", show_label=False, max_lines=1, elem_id="prompt_input"
                    ).style(
                        border=(True, False, True, True),
                        rounded=(True, False, False, True),
                        container=False                        
                    )
                    btn = gr.Button("Run",elem_id="run_btn").style(
                        margin=False,
                        rounded=(False, True, True, False)
                    )  
                with gr.Row().style():
                    infer_outputs = gr.Gallery(show_label=False).style(grid=[2], height="512px")
                with gr.Row():
                  gr.HTML("<p style=\"font-size: 85%;margin-top: .75em\">Prompting may not work as you are used to; <code>objects</code> may need the concept added at the end.</p>")
                with gr.Row():
                  gr.Examples(examples=examples, fn=infer, inputs=[text], outputs=infer_outputs, cache_examples=False)
  checkbox_states = {}
  inputs = [text]
  btn.click(
        infer,
        inputs=inputs,
        outputs=infer_outputs
    )
demo.launch(inline=False, debug=True)