File size: 10,083 Bytes
ac6bfa9
 
 
 
 
 
 
 
 
 
 
 
 
b329690
ac6bfa9
 
 
 
 
 
 
 
 
 
 
 
386d492
ac6bfa9
 
 
 
 
 
 
b329690
 
 
ac6bfa9
 
 
 
 
 
 
 
 
 
 
 
 
 
 
62ee15a
ac6bfa9
 
 
 
 
b329690
 
bd2291f
ac6bfa9
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
bd2291f
ac6bfa9
 
 
 
62ee15a
ac6bfa9
 
 
 
 
 
 
b329690
 
 
 
ac6bfa9
 
b329690
 
 
 
 
 
 
ac6bfa9
 
 
 
179908e
ac6bfa9
 
 
 
 
 
 
b329690
ac6bfa9
 
b329690
ac6bfa9
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
b329690
 
 
 
 
 
 
 
 
 
 
 
ac6bfa9
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6732512
 
 
 
b329690
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6732512
 
ac6bfa9
 
 
 
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
import streamlit as st

import torch
from transformers import pipeline, set_seed
from transformers import AutoTokenizer

from PIL import (
    Image,
    ImageFont,
    ImageDraw
)

import re
import textwrap
from examples import EXAMPLES
import meta
from utils import (
    remote_css,
    local_css,
    load_image,
    pure_comma_separation
)


class TextGeneration:
    def __init__(self):
        self.debug = False
        self.dummy_output = {
            'directions': [
                'peel the potato and slice thinly.',
                'place in a microwave safe dish.',
                'cover with plastic wrap and microwave on high for 5 minutes.',
                'remove from the microwave and sprinkle with cheese.',
                'return to the microwave for 1 minute or until cheese is melted.',
                'return to the microwave for 1 minute or until cheese is melted. return to the microwave for 1 minute or until cheese is melted.'
                'return to the microwave for 1 minute or until cheese is melted.',
                'return to the microwave for 1 minute or until cheese is melted.',
                'return to the microwave for 1 minute or until cheese is melted.',
            ],
            'ingredients': [
                '1 potato',
                '1 slice cheese',
                '1 potato',
                '1 slice cheese'
                '1 potato',
                '1 slice cheese',
                '1 slice cheese',
                '1 potato',
                '1 slice cheese'
                '1 potato',
                '1 slice cheese',
            ],
            'title': 'Cheese Potatoes with Some other items'
        }
        self.tokenizer = None
        self.generator = None
        self.task = "text2text-generation"
        self.model_name_or_path = "flax-community/t5-recipe-generation"
        self.list_division = 5
        self.point = "-"
        self.h1_font = ImageFont.truetype("asset/fonts/PT_Serif/PTSerif-Bold.ttf", 75)
        self.h2_font = ImageFont.truetype("asset/fonts/PT_Serif/PTSerif-Bold.ttf", 50)
        self.p_font = ImageFont.truetype("asset/fonts/PT_Serif/PTSerif-Regular.ttf", 30)
        set_seed(42)

    def _skip_special_tokens_and_prettify(self, text):
        recipe_maps = {"<sep>": "--", "<section>": "\n"}
        recipe_map_pattern = "|".join(map(re.escape, recipe_maps.keys()))

        text = re.sub(
            recipe_map_pattern,
            lambda m: recipe_maps[m.group()],
            re.sub("|".join(self.tokenizer.all_special_tokens), "", text)
        )

        data = {"title": "", "ingredients": [], "directions": []}
        for section in text.split("\n"):
            section = section.strip()
            if section.startswith("title:"):
                data["title"] = " ".join(
                    [w.strip().capitalize() for w in section.replace("title:", "").strip().split() if w.strip()]
                )
            elif section.startswith("ingredients:"):
                data["ingredients"] = [s.strip() for s in section.replace("ingredients:", "").split('--')]
            elif section.startswith("directions:"):
                data["directions"] = [s.strip() for s in section.replace("directions:", "").split('--')]
            else:
                pass

        return data

    def load(self):
        if not self.debug:
            self.tokenizer = AutoTokenizer.from_pretrained(self.model_name_or_path)
            self.generator = pipeline(self.task, model=self.model_name_or_path, tokenizer=self.model_name_or_path)

    def prepare_frame(self, recipe, frame):
        im_editable = ImageDraw.Draw(frame)
        # Title
        ws, hs = 120, 500
        im_editable.text(
            (ws, hs),
            textwrap.fill(recipe["title"], 25),
            (61, 61, 70),
            font=self.h1_font,
        )
        # Ingredients
        hs = hs + 270
        im_editable.text(
            (ws, hs),
            "Ingredients",
            (61, 61, 70),
            font=self.h2_font,
        )
        hs = hs + 80
        ingredients = recipe["ingredients"]
        ingredients_col1 = [textwrap.fill(item, 30) for item in ingredients[:self.list_division]]
        ingredients_col2 = [textwrap.fill(item, 30) for item in ingredients[self.list_division:]]

        im_editable.text(
            (ws + 10, hs),
            "\n".join([f"- {item}" for item in ingredients_col1]),
            (61, 61, 70),
            font=self.p_font,
        )
        im_editable.text(
            (ws + 500, hs),
            "\n".join([f"{self.point} {item}" for item in ingredients_col2]),
            (61, 61, 70),
            font=self.p_font,
        )
        # Directions
        hs = hs + 200
        im_editable.text(
            (ws, hs),
            "Directions",
            (61, 61, 70),
            font=self.h2_font,
        )
        hs = hs + 80
        directions = [textwrap.fill(item, 70).replace("\n", "\n   ") for item in recipe["directions"]]
        im_editable.text(
            (ws + 10, hs),
            "\n".join([f"{self.point} {item}" for item in directions]),
            (61, 61, 70),
            font=self.p_font,
        )
        return frame

    def generate(self, items, generation_kwargs):
        print(generation_kwargs)
        if not self.debug:
            generation_kwargs["num_return_sequences"] = 1
            # generation_kwargs["return_full_text"] = False
            generation_kwargs["return_tensors"] = True
            generation_kwargs["return_text"] = False

            generated_ids = self.generator(
                items,
                **generation_kwargs,
            )[0]["generated_token_ids"]
            recipe = self.tokenizer.decode(generated_ids, skip_special_tokens=False)
            recipe = self._skip_special_tokens_and_prettify(recipe)
            return recipe

        return self.dummy_output

    def generate_frame(self, recipe):
        frame = load_image("asset/images/recipe-post.png")
        return self.prepare_frame(recipe, frame)


@st.cache(allow_output_mutation=True)
def load_text_generator():
    generator = TextGeneration()
    generator.load()
    return generator


chef_top = {
    "max_length": 512,
    "min_length": 64,
    "no_repeat_ngram_size": 3,
    "do_sample": True,
    "top_k": 60,
    "top_p": 0.95,
    "num_return_sequences": 1
}
chef_beam = {
    "max_length": 512,
    "min_length": 64,
    "no_repeat_ngram_size": 3,
    "early_stopping": True,
    "num_beams": 5,
    "length_penalty": 1.5,
    "num_return_sequences": 1
}


def main():
    st.set_page_config(
        page_title="Chef Transformer",
        page_icon="🍲",
        layout="wide",
        initial_sidebar_state="expanded"
    )
    generator = load_text_generator()
    local_css("asset/css/style.css")

    st.sidebar.image(load_image("asset/images/chef-transformer-transparent.png"), width=310)
    st.sidebar.title("Welcome to our lovely restaurant, what can I do for you?")
    chef = st.sidebar.selectbox("Choose your chef", index=0, options=["Chef Scheherazade", "Chef Giovanni"])
    is_text = st.sidebar.selectbox(
        label='Recipe',
        options=(True, False),
        help="Will generate your recipe as a text post",
    )
    is_frame = st.sidebar.selectbox(
        label='Recipe for Instagram?',
        options=(True, False),
        help="Will generate your recipe as an Instagram post",
    )

    st.markdown(meta.HEADER_INFO)
    prompts = list(EXAMPLES.keys()) + ["Custom"]
    prompt = st.selectbox('Examples', prompts, index=len(prompts) - 1)

    if prompt == "Custom":
        prompt_box = ""
    else:
        prompt_box = EXAMPLES[prompt]

    items = st.text_input(
        'Add custom ingredients here (separated by `,`): ',
        pure_comma_separation(prompt_box, return_list=False),
        key="custom_keywords",
        max_chars=1000)
    items = pure_comma_separation(items, return_list=False)
    entered_items = st.empty()

    if st.button('Get Recipe!'):
        entered_items.markdown("**Generate recipe for:** " + items)
        with st.spinner("Generating recipe..."):

            if isinstance(items, str) and len(items) > 1:
                gen_kw = chef_top if chef == "Chef Scheherazade" else chef_beam
                generated_recipe = generator.generate(items, gen_kw)

                if is_text:
                    title = generated_recipe["title"]
                    ingredients = generated_recipe["ingredients"]
                    directions = [textwrap.fill(item, 70).replace("\n", "\n   ") for item in
                                  generated_recipe["directions"]]
                    st.markdown(
                        " ".join([
                            f"<h2>{title}</h2>",
                            "<h3>Ingredient</h3>",
                            "<ul class='ingredients-list'>",
                            " ".join([f'<li>{item}</li>' for item in ingredients]),
                            "</ul>",
                            "<h3>Direction</h3>",
                            "<ul class='ingredients-list'>",
                            " ".join([f'<li>{item}</li>' for item in directions]),
                            "</ul>",
                        ]),
                        unsafe_allow_html=True
                    )

                if is_frame:
                    recipe_post = generator.generate_frame(generated_recipe)

                    col1, col2, col3 = st.beta_columns([1, 6, 1])
                    with col1:
                        st.write("")

                    with col2:
                        st.image(
                            recipe_post,
                            # width=500,
                            caption="Your recipe",
                            use_column_width="auto",
                            output_format="PNG"
                        )

                    with col3:
                        st.write("")
            else:
                entered_items.markdown("Enter your items...")


if __name__ == '__main__':
    main()