File size: 17,957 Bytes
6b5dfe6
bfdbdf6
 
 
 
 
182990e
b3d3b2f
666a605
 
ded3b8b
6b5dfe6
 
 
 
 
b3d3b2f
bfdbdf6
 
 
 
 
 
 
c4b99ca
bfdbdf6
 
 
c4b99ca
bfdbdf6
c4b99ca
bfdbdf6
c4b99ca
bfdbdf6
ac586a8
 
666a605
ac586a8
 
666a605
 
 
 
ac586a8
 
 
 
 
 
 
 
 
666a605
c24dac7
666a605
 
 
 
 
c9de947
 
bfdbdf6
 
 
 
 
 
 
666a605
 
b3d3b2f
 
 
 
 
 
 
 
 
 
 
ded3b8b
22d5d2c
bfdbdf6
 
9255bd7
bfdbdf6
d1c3953
bfdbdf6
 
 
9465fd2
bfdbdf6
666a605
 
 
 
 
 
 
 
 
 
d1c3953
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
666a605
 
126bc9f
cb024a4
666a605
c9de947
ded3b8b
ac586a8
c9de947
 
 
 
 
 
 
 
 
666a605
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
d1c3953
ac586a8
cb024a4
 
 
6b5dfe6
 
c4b99ca
 
 
 
 
 
 
 
 
ac586a8
 
 
 
666a605
ac586a8
 
6b5dfe6
bfdbdf6
6b5dfe6
bfdbdf6
c4b99ca
6b5dfe6
 
bfdbdf6
 
c4b99ca
bfdbdf6
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6b5dfe6
bfdbdf6
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
666a605
 
 
bfdbdf6
 
b3d3b2f
62bbb3e
ac586a8
666a605
 
c24dac7
666a605
bfdbdf6
9255bd7
666a605
9255bd7
 
666a605
9255bd7
 
666a605
 
 
 
 
9255bd7
ac586a8
666a605
ac586a8
cb024a4
ac586a8
666a605
 
ac586a8
6b5dfe6
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
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
import gradio as gr
import os
from pathlib import Path
import argparse
import shutil
from train_dreambooth import run_training
from convertosd import convert
from PIL import Image
from slugify import slugify
import requests
import torch
css = '''
    .instruction{position: absolute; top: 0;right: 0;margin-top: 0px !important}
    .arrow{position: absolute;top: 0;right: -8px;margin-top: -8px !important}
    #component-4, #component-3, #component-10{min-height: 0}
'''
model_to_load = "multimodalart/sd-fine-tunable"
maximum_concepts = 3

def swap_text(option):
    mandatory_liability = "You must have the right to do so and you are liable for the images you use"
    if(option == "object"):
        instance_prompt_example = "cttoy"
        freeze_for = 50
        return [f"You are going to train `object`(s), upload 5-10 images of each object you are planning on training on from different angles/perspectives. {mandatory_liability}:", '''<img src="file/cat-toy.png" />''', f"You should name your concept with a unique made up word that has low chance of the model already knowing it (e.g.: `{instance_prompt_example}` here). Images will be automatically cropped to 512x512.", freeze_for]
    elif(option == "person"):
       instance_prompt_example = "julcto"
       freeze_for = 100
       return [f"You are going to train a `person`(s), upload 10-20 images of each person you are planning on training on from different angles/perspectives. {mandatory_liability}:", '''<img src="file/person.png" />''', f"You should name the files with a unique word that represent your concept (e.g.: `{instance_prompt_example}` here). Images will be automatically cropped to 512x512.", freeze_for]
    elif(option == "style"):
        instance_prompt_example = "trsldamrl"
        freeze_for = 10
        return [f"You are going to train a `style`, upload 10-20 images of the style you are planning on training on. Name the files with the words you would like  {mandatory_liability}:", '''<img src="file/trsl_style.png" />''', f"You should name your files with a unique word that represent your concept (e.g.: `{instance_prompt_example}` here). Images will be automatically cropped to 512x512.", freeze_for]

def count_files(*inputs):
    file_counter = 0
    concept_counter = 0
    for i, input in enumerate(inputs):
        if(i < maximum_concepts-1):
            files = inputs[i]
            if(files):
                concept_counter+=1
                file_counter+=len(files)
    uses_custom = inputs[-1] 
    type_of_thing = inputs[-4]
    if(uses_custom):
        Training_Steps = int(inputs[-3])
    else:
        if(type_of_thing == "person"):
            Training_Steps = file_counter*200*2
        else:
            Training_Steps = file_counter*200
    return(gr.update(visible=True, value=f"You are going to train {concept_counter} {type_of_thing}(s), with {file_counter} images for {Training_Steps} steps. This should take around {round(Training_Steps/1.5, 2)} seconds, or {round((Training_Steps/1.5)/3600, 2)} hours. As a reminder, the T4 GPU costs US$0.60 for 1h. Once training is over, don't forget to swap the hardware back to CPU."))

def train(*inputs):
    if "IS_SHARED_UI" in os.environ:
        raise gr.Error("This Space only works in duplicated instances")
    if os.path.exists("output_model"): shutil.rmtree('output_model')
    if os.path.exists("instance_images"): shutil.rmtree('instance_images')
    if os.path.exists("diffusers_model.zip"): os.remove("diffusers_model.zip")
    if os.path.exists("model.ckpt"): os.remove("model.ckpt")
    file_counter = 0
    for i, input in enumerate(inputs):
        if(i < maximum_concepts-1):
            if(input):
                os.makedirs('instance_images',exist_ok=True)
                files = inputs[i+(maximum_concepts*2)]
                prompt = inputs[i+maximum_concepts]
                if(prompt == "" or prompt == None):
                    raise gr.Error("You forgot to define your concept prompt")
                for j, file_temp in enumerate(files):
                    file = Image.open(file_temp.name)
                    width, height = file.size
                    side_length = min(width, height)
                    left = (width - side_length)/2
                    top = (height - side_length)/2
                    right = (width + side_length)/2
                    bottom = (height + side_length)/2
                    image = file.crop((left, top, right, bottom))
                    image = image.resize((512, 512))
                    extension = file_temp.name.split(".")[1]
                    image = image.convert('RGB')
                    image.save(f'instance_images/{prompt}_({j+1}).jpg', format="JPEG", quality = 100)
                    file_counter += 1
    
    os.makedirs('output_model',exist_ok=True)
    uses_custom = inputs[-1] 
    type_of_thing = inputs[-4]
    if(uses_custom):
        Training_Steps = int(inputs[-3])
        Train_text_encoder_for = int(inputs[-2])
    else:
        Training_Steps = file_counter*200
        if(type_of_thing == "object"):
            Train_text_encoder_for=30
        elif(type_of_thing == "person"):
            Train_text_encoder_for=60
        elif(type_of_thing == "style"):
            Train_text_encoder_for=15
    
    class_data_dir = None
    stptxt = int((Training_Steps*Train_text_encoder_for)/100)
    args_general = argparse.Namespace(
                image_captions_filename = True,
                train_text_encoder = True,
                stop_text_encoder_training = stptxt,
                save_n_steps = 0,
                pretrained_model_name_or_path = model_to_load,
                instance_data_dir="instance_images",
                class_data_dir=class_data_dir,
                output_dir="output_model",
                instance_prompt="",
                seed=42,
                resolution=512,
                mixed_precision="fp16",
                train_batch_size=1,
                gradient_accumulation_steps=1,
                use_8bit_adam=True,
                learning_rate=2e-6,
                lr_scheduler="polynomial",
                lr_warmup_steps = 0,
                max_train_steps=Training_Steps,     
    )
    run_training(args_general)
    torch.cuda.empty_cache()
    #convert("output_model", "model.ckpt")
    #shutil.rmtree('instance_images')
    shutil.make_archive("diffusers_model", 'zip', "output_model")
    torch.cuda.empty_cache()
    return [gr.update(visible=True, value=["diffusers_model.zip"]), gr.update(visible=True), gr.update(visible=True), gr.update(visible=True)]

def generate(prompt):
    from diffusers import StableDiffusionPipeline
    
    pipe = StableDiffusionPipeline.from_pretrained("./output_model", torch_dtype=torch.float16)
    pipe = pipe.to("cuda")
    image = pipe(prompt).images[0]  
    return(image)
    
def push(model_name, where_to_upload, hf_token):
    if(not os.path.exists("model.ckpt")):
        convert("output_model", "model.ckpt")
    from huggingface_hub import HfApi, HfFolder, CommitOperationAdd
    from huggingface_hub import create_repo
    model_name_slug = slugify(model_name)
    if(where_to_upload == "My personal profile"):
        api = HfApi()
        your_username = api.whoami(token=hf_token)["name"]
        model_id = f"{your_username}/{model_name_slug}"
    else:
        model_id = f"sd-dreambooth-library/{model_name_slug}"
        headers = {"Authorization" : f"Bearer: {hf_token}", "Content-Type": "application/json"}
        response = requests.post("https://example.com/get-my-account-detail", headers=headers)
    
    images_upload = os.listdir("instance_images")
    image_string = ""
    instance_prompt_list = []
    previous_instance_prompt = ''
    for i, image in enumerate(images_upload):
        instance_prompt = image.split("_")[0]
        if(instance_prompt != previous_instance_prompt):
            title_instance_prompt_string = instance_prompt
            instance_prompt_list.append(instance_prompt)
        else:
            title_instance_prompt_string = ''
        previous_instance_prompt = instance_prompt
        image_string = f'''
    {title_instance_prompt_string}
    {image_string}![{instance_prompt} {i}](https://huggingface.co/{model_name_slug}/resolve/main/sample_images/{image})
        '''
    readme_text = f'''---
license: creativeml-openrail-m
tags:
- text-to-image
---
### {model_name} Dreambooth model trained by {api.whoami(token=hf_token)["name"]} with [Hugging Face Dreambooth Training Space](https://huggingface.co/spaces/multimodalart/dreambooth-training)

You run your new concept via `diffusers` [Colab Notebook for Inference](https://colab.research.google.com/github/huggingface/notebooks/blob/main/diffusers/sd_dreambooth_inference.ipynb)

Sample pictures of this concept:
{image_string}
'''
    #Save the readme to a file
    readme_file = open("README.md", "w")
    readme_file.write(readme_text)
    readme_file.close()
    #Save the token identifier to a file
    text_file = open("token_identifier.txt", "w")
    text_file.write(', '.join(instance_prompt_list))
    text_file.close()
    operations = [
        CommitOperationAdd(path_in_repo="token_identifier.txt", path_or_fileobj="token_identifier.txt"),
        CommitOperationAdd(path_in_repo="README.md", path_or_fileobj="README.md"),
        CommitOperationAdd(path_in_repo=f"model.ckpt",path_or_fileobj="model.ckpt")
    ]
    api.create_commit(
    repo_id=model_id,
    operations=operations,
    commit_message=f"Upload the model {model_name}",
    token=hf_token
    )
    api.upload_folder(
    folder_path="output_model",
    repo_id=model_id,
    token=hf_token
    )
    api.upload_folder(
    folder_path="instance_images",
    path_in_repo="concept_images",
    repo_id=model_id,
    token=hf_token
    )
    return [gr.update(visible=True, value=f"Successfully uploaded your model. Access it [here](https://huggingface.co/{model_id})"), gr.update(visible=True, value=["diffusers_model.zip", "model.ckpt"])]

def convert_to_ckpt():
    convert("output_model", "model.ckpt")
    return gr.update(visible=True, value=["diffusers_model.zip", "model.ckpt"])
   
with gr.Blocks(css=css) as demo:
    with gr.Box():
        if "IS_SHARED_UI" in os.environ:
            gr.HTML('''
                <div class="gr-prose" style="max-width: 80%">
                <h2>Attention - This Space doesn't work in this shared UI</h2>
                <p>For it to work, you have to duplicate the Space and run it on your own profile where a (paid) private GPU will be attributed to it during runtime. It will cost you < US$1 to train a model on default settings! 🤑</p> 
                <img class="instruction" src="file/duplicate.png"> 
                <img class="arrow" src="file/arrow.png" />
                </div>
            ''')
        else:
            gr.HTML('''
                <div class="gr-prose" style="max-width: 80%">
                <h2>You have successfully cloned the Dreambooth Training Space</h2>
                <p>If you haven't already, attribute a T4 GPU to it (via the Settings tab) and run the training below. You will be billed by the minute between when you activate the GPU until when you turn it off.</p> 
                </div>
            ''')    
    gr.Markdown("# Dreambooth training")
    gr.Markdown("Customize Stable Diffusion by giving it with few-shot examples")
    with gr.Row():
        type_of_thing = gr.Dropdown(label="What would you like to train?", choices=["object", "person", "style"], value="object", interactive=True)
       
    with gr.Row():
        with gr.Column():
            thing_description = gr.Markdown("You are going to train an `object`, upload 5-10 images of the object you are planning on training on from different angles/perspectives. You must have the right to do so and you are liable for the images you use")
            thing_image_example = gr.HTML('''<img src="file/cat-toy.png" />''')
            things_naming = gr.Markdown("You should name your concept with a unique made up word that has low chance of the model already knowing it (e.g.: `cttoy` here). Images will be automatically cropped to 512x512.")
        with gr.Column():
            file_collection = []
            concept_collection = []
            buttons_collection = []
            delete_collection = []
            is_visible = []

            row = [None] * maximum_concepts
            for x in range(maximum_concepts):
                ordinal = lambda n: "%d%s" % (n, "tsnrhtdd"[(n // 10 % 10 != 1) * (n % 10 < 4) * n % 10::4])
                if(x == 0):
                    visible = True
                    is_visible.append(gr.State(value=True))
                else:
                    visible = False
                    is_visible.append(gr.State(value=False))

                file_collection.append(gr.File(label=f"Upload the images for your {ordinal(x+1)} concept", file_count="multiple", interactive=True, visible=visible))
                with gr.Column(visible=visible) as row[x]:
                    concept_collection.append(gr.Textbox(label=f"{ordinal(x+1)} concept prompt - use a unique, made up word to avoid collisions"))  
                    with gr.Row():
                        if(x < maximum_concepts-1):
                            buttons_collection.append(gr.Button(value="Add +1 concept", visible=visible))
                        if(x > 0):
                            delete_collection.append(gr.Button(value=f"Delete {ordinal(x+1)} concept"))
            
            counter_add = 1
            for button in buttons_collection:
                if(counter_add < len(buttons_collection)):
                    button.click(lambda:
                    [gr.update(visible=True),gr.update(visible=True), gr.update(visible=False), gr.update(visible=True), True, None],
                    None, 
                    [row[counter_add], file_collection[counter_add], buttons_collection[counter_add-1], buttons_collection[counter_add], is_visible[counter_add], file_collection[counter_add]])
                else:
                    button.click(lambda:[gr.update(visible=True),gr.update(visible=True), gr.update(visible=False), True], None, [row[counter_add], file_collection[counter_add], buttons_collection[counter_add-1], is_visible[counter_add]])
                counter_add += 1
            
            counter_delete = 1
            for delete_button in delete_collection:
                if(counter_delete < len(delete_collection)+1):
                    delete_button.click(lambda:[gr.update(visible=False),gr.update(visible=False), gr.update(visible=True), False], None, [file_collection[counter_delete], row[counter_delete], buttons_collection[counter_delete-1], is_visible[counter_delete]])
                counter_delete += 1
            
            
            
    with gr.Accordion("Custom Settings", open=False):
        swap_auto_calculated = gr.Checkbox(label="Use custom settings")
        gr.Markdown("If not checked, the number of steps and % of frozen encoder will be tuned automatically according to the amount of images you upload and whether you are training an `object`, `person` or `style` as follows: The number of steps is calculated by number of images uploaded multiplied by 20. The text-encoder is frozen after 10% of the steps for a style, 30% of the steps for an object and is fully trained for persons.")
        steps = gr.Number(label="How many steps", value=800)
        perc_txt_encoder = gr.Number(label="Percentage of the training steps the text-encoder should be trained as well", value=30)

    type_of_thing.change(fn=swap_text, inputs=[type_of_thing], outputs=[thing_description, thing_image_example, things_naming, perc_txt_encoder], queue=False)
    training_summary = gr.Textbox("", visible=False, label="Training Summary")
    steps.change(fn=count_files, inputs=file_collection+[type_of_thing]+[steps]+[perc_txt_encoder]+[swap_auto_calculated], outputs=[training_summary])
    perc_txt_encoder.change(fn=count_files, inputs=file_collection+[type_of_thing]+[steps]+[perc_txt_encoder]+[swap_auto_calculated], outputs=[training_summary])
    for file in file_collection:
        file.change(fn=count_files, inputs=file_collection+[type_of_thing]+[steps]+[perc_txt_encoder]+[swap_auto_calculated], outputs=[training_summary])
    train_btn = gr.Button("Start Training")
    with gr.Box(visible=False) as try_your_model:
        gr.Markdown("## Try your model")
        with gr.Row():
            prompt = gr.Textbox(label="Type your prompt")
            result_image = gr.Image()
        generate_button = gr.Button("Generate Image")
    with gr.Box(visible=False) as push_to_hub:
        gr.Markdown("## Push to Hugging Face Hub")
        model_name = gr.Textbox(label="Name of your model", placeholder="Tarsila do Amaral Style")
        where_to_upload = gr.Dropdown(["My personal profile", "Public Library"], label="Upload to")
        gr.Markdown("[A Hugging Face write access token](https://huggingface.co/settings/tokens), go to \"New token\" -> Role : Write. A regular read token won't work here.")
        hf_token = gr.Textbox(label="Hugging Face Write Token")
        push_button = gr.Button("Push to the Hub")
    result = gr.File(label="Download the uploaded models in the diffusers format", visible=True)
    success_message_upload = gr.Markdown(visible=False)
    convert_button = gr.Button("Convert to CKPT", visible=False)

    train_btn.click(fn=train, inputs=is_visible+concept_collection+file_collection+[type_of_thing]+[steps]+[perc_txt_encoder]+[swap_auto_calculated], outputs=[result, try_your_model, push_to_hub, convert_button])
    generate_button.click(fn=generate, inputs=prompt, outputs=result_image)
    push_button.click(fn=push, inputs=[model_name, where_to_upload, hf_token], outputs=[success_message_upload, result])
    convert_button.click(fn=convert_to_ckpt, inputs=[], outputs=result)
demo.launch()