File size: 4,531 Bytes
a0b5629
 
 
 
 
 
 
 
 
8413cb1
a0b5629
 
 
 
 
 
 
 
 
 
8413cb1
 
03410c0
8413cb1
 
03410c0
a0b5629
 
03410c0
 
 
 
 
33669c4
03410c0
 
 
 
 
 
 
 
33669c4
03410c0
 
33669c4
03410c0
 
33669c4
a0b5629
 
03410c0
a0b5629
 
f771a11
a0b5629
ce875ce
03410c0
 
ede6cee
4032f9e
03410c0
 
 
4032f9e
a0b5629
 
 
 
03410c0
ede6cee
 
33669c4
03410c0
a396b07
a0b5629
 
 
03410c0
a0b5629
 
 
 
 
 
 
 
 
 
03410c0
33669c4
a0b5629
 
 
 
 
 
 
 
 
 
 
 
 
 
33669c4
a0b5629
 
 
03410c0
a0b5629
 
8413cb1
 
fa7a46e
03410c0
a0b5629
 
 
 
 
8413cb1
a396b07
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
import gradio as gr
import re
import requests
import json
import os
from screenshot import BG_COMP, BOX_COMP, GENERATION_VAR, PROMPT_VAR, main
from pathlib import Path

title = "BLOOM"

description = """Gradio Demo for BLOOM. To use it, simply add your text, or click one of the examples to load them.
Tips:
- Do NOT talk to BLOOM as an entity, it's not a chatbot but a webpage/blog/article completion model.
- For the best results: MIMIC a few sentences of a webpage similar to the content you want to generate.
Start a paragraph as if YOU were writing a blog, webpage, math post, coding article and BLOOM will generate a coherent follow-up. Longer prompts usually give more interesting results.
Options:
- sampling: imaginative completions (may be not super accurate e.g. math/history)
- greedy: accurate completions (may be more boring or have repetitions)
"""

wip_description = """JAX / Flax Gradio Demo for BLOOM. The 176B BLOOM model running on a TPU v3-256 pod, with 2D model parallelism and custom mesh axes.

Note: for this WIP demo only **sampling** is supported.
"""

API_URL = os.getenv("API_URL")

examples = [
    [
        'A "whatpu" is a small, furry animal native to Tanzania. An example of a sentence that uses the word whatpu is: We were traveling in Africa and we saw these very cute whatpus. To do a "farduddle" means to jump up and down really fast. An example of a sentence that uses the word farduddle is:',
        64, "sampling", False],
    ['A poem about the beauty of science by Alfred Edgar Brittle\nTitle: The Magic Craft\nIn the old times', 64,
     "sampling", False],
    ['استخراج العدد العاملي في لغة بايثون:', 64, "sampling", False],
    ["Pour déguster un ortolan, il faut tout d'abord", 64, "sampling", False],
    [
        'Traduce español de España a español de Argentina\nEl coche es rojo - el auto es rojo\nEl ordenador es nuevo - la computadora es nueva\nel boligrafo es negro -',
        64, "sampling", False],
    [
        'Estos ejemplos quitan vocales de las palabras\nEjemplos:\nhola - hl\nmanzana - mnzn\npapas - pps\nalacran - lcrn\npapa -',
        64, "sampling", False],
    ["Question: If I put cheese into the fridge, will it melt?\nAnswer:", 64, "sampling", False],
    ["Math exercise - answers:\n34+10=44\n54+20=", 64, "sampling", False],
    [
        "Question: Where does the Greek Goddess Persephone spend half of the year when she is not with her mother?\nAnswer:",
        64, "sampling", False],
    [
        "spelling test answers.\nWhat are the letters in « language »?\nAnswer: l-a-n-g-u-a-g-e\nWhat are the letters in « Romanian »?\nAnswer:",
        64, "sampling", False]
]


def query(payload):
    print(payload)
    response = requests.post(API_URL, json=payload)
    print(response)
    return response.json()


def inference(input_sentence, max_length, sample_or_greedy, raw_text=True):
    payload = {
        "prompt": input_sentence,
        "do_sample": True,
        # "max_new_tokens": max_length
    }

    data = query(
        payload
    )

    # if raw_text:
    if True:
        return None, data[0]['generated_text']

    width, height = 3326, 3326
    assets_path = "assets"
    font_mapping = {
        "latin characters (faster)": "DejaVuSans.ttf",
        "complete alphabet (slower)": "GoNotoCurrent.ttf"
    }
    working_dir = Path(__file__).parent.resolve()
    font_path = str(working_dir / font_mapping["complete alphabet (slower)"])
    img_save_path = str(working_dir / "output.jpeg")
    colors = {
        BG_COMP: "#000000",
        PROMPT_VAR: "#FFFFFF",
        GENERATION_VAR: "#FF57A0",
        BOX_COMP: "#120F25",
    }

    new_string = data[0]['generated_text']

    _, img = main(
        input_sentence,
        new_string,
        width,
        height,
        assets_path=assets_path,
        font_path=font_path,
        colors=colors,
        frame_to_box_margin=200,
        text_to_text_box_margin=50,
        init_font_size=150,
        right_align=False,
    )
    return img, data[0]['generated_text']


gr.Interface(
    inference,
    [
        gr.inputs.Textbox(label="Input"),
        gr.inputs.Radio([64], default=64, label="Tokens to generate"),
        gr.inputs.Radio(["sampling"], label="Sample or greedy", default="sampling"),
        gr.Checkbox(label="Just output raw text", value=True),
    ],
    ["image", "text"],
    examples=examples,
    # article=article,
    cache_examples=False,
    title=title,
    description=wip_description
).launch()