Spaces:
Sleeping
Sleeping
liujch1998
commited on
Commit
β’
25f66ac
1
Parent(s):
ab4c12e
Initial commit
Browse files- README.md +4 -4
- app.py +127 -0
- constants.py +24 -0
- requirements.txt +6 -0
README.md
CHANGED
@@ -1,8 +1,8 @@
|
|
1 |
---
|
2 |
-
title: Creativity
|
3 |
-
emoji:
|
4 |
-
colorFrom:
|
5 |
-
colorTo:
|
6 |
sdk: gradio
|
7 |
sdk_version: 4.42.0
|
8 |
app_file: app.py
|
|
|
1 |
---
|
2 |
+
title: Creativity Index
|
3 |
+
emoji: π©π½βπ¨
|
4 |
+
colorFrom: blue
|
5 |
+
colorTo: green
|
6 |
sdk: gradio
|
7 |
sdk_version: 4.42.0
|
8 |
app_file: app.py
|
app.py
ADDED
@@ -0,0 +1,127 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
import datetime
|
3 |
+
import json
|
4 |
+
import requests
|
5 |
+
from constants import *
|
6 |
+
|
7 |
+
def process(query_type, index_desc, **kwargs):
|
8 |
+
timestamp = datetime.datetime.now().strftime('%Y%m%d-%H%M%S')
|
9 |
+
index = INDEX_BY_DESC[index_desc]
|
10 |
+
data = {
|
11 |
+
'source': 'hf' if not DEBUG else 'hf-dev',
|
12 |
+
'timestamp': timestamp,
|
13 |
+
'query_type': query_type,
|
14 |
+
'index': index,
|
15 |
+
}
|
16 |
+
data.update(kwargs)
|
17 |
+
print(json.dumps(data))
|
18 |
+
if API_URL is None:
|
19 |
+
raise ValueError(f'API_URL envvar is not set!')
|
20 |
+
try:
|
21 |
+
response = requests.post(API_URL, json=data, timeout=10)
|
22 |
+
except requests.exceptions.Timeout:
|
23 |
+
raise ValueError('Web request timed out. Please try again later.')
|
24 |
+
except requests.exceptions.RequestException as e:
|
25 |
+
raise ValueError(f'Web request error: {e}')
|
26 |
+
if response.status_code == 200:
|
27 |
+
result = response.json()
|
28 |
+
else:
|
29 |
+
raise ValueError(f'HTTP error {response.status_code}: {response.json()}')
|
30 |
+
if DEBUG:
|
31 |
+
print(result)
|
32 |
+
return result
|
33 |
+
|
34 |
+
def creativity(index_desc, query):
|
35 |
+
result = process('creativity', index_desc, query=query)
|
36 |
+
latency = '' if 'latency' not in result else f'{result["latency"]:.3f}'
|
37 |
+
if 'error' in result:
|
38 |
+
ci = result['error']
|
39 |
+
ngram_len = NGRAM_LEN_DEFAULT
|
40 |
+
html = ''
|
41 |
+
return latency, ci, ngram_len, html
|
42 |
+
|
43 |
+
rs = result['rs']
|
44 |
+
tokens = result['tokens']
|
45 |
+
highlighteds_by_n = {}
|
46 |
+
uniqueness_by_n = {}
|
47 |
+
for n in range(NGRAM_LEN_MIN, NGRAM_LEN_MAX + 1):
|
48 |
+
highlighteds = [False] * len(tokens)
|
49 |
+
last_r = 0
|
50 |
+
for l, r in enumerate(rs):
|
51 |
+
if r - l < n:
|
52 |
+
continue
|
53 |
+
for i in range(max(last_r, l), r):
|
54 |
+
highlighteds[i] = True
|
55 |
+
last_r = r
|
56 |
+
uniqueness = sum([1 for h in highlighteds if not h]) / len(highlighteds)
|
57 |
+
highlighteds_by_n[n] = highlighteds
|
58 |
+
uniqueness_by_n[n] = uniqueness
|
59 |
+
ci = sum(uniqueness_by_n.values()) / len(uniqueness_by_n)
|
60 |
+
ci = f'{ci:.2%}'
|
61 |
+
|
62 |
+
ngram_len = NGRAM_LEN_DEFAULT
|
63 |
+
|
64 |
+
html = ''
|
65 |
+
highlighted = highlighteds_by_n[ngram_len]
|
66 |
+
line_len = 0
|
67 |
+
for i, (token, highlighted) in enumerate(zip(tokens, highlighteds)):
|
68 |
+
if line_len >= 100 and token.startswith('Δ ') and token != 'Δ':
|
69 |
+
html += '<br/>'
|
70 |
+
line_len = 0
|
71 |
+
color = '0, 0, 255, 0.5'
|
72 |
+
if token == 'Δ':
|
73 |
+
disp_token = '\\n'
|
74 |
+
is_linebreak = True
|
75 |
+
else:
|
76 |
+
disp_token = token.replace('Δ ', ' ')
|
77 |
+
is_linebreak = False
|
78 |
+
if highlighted:
|
79 |
+
html += f'<span id="hldoc-token-{i}" style="background-color: rgba{color};" class="background-color: rgba{color};">{disp_token}</span>'
|
80 |
+
else:
|
81 |
+
html += disp_token
|
82 |
+
if is_linebreak:
|
83 |
+
html += '<br/>'
|
84 |
+
line_len = 0
|
85 |
+
else:
|
86 |
+
line_len += len(token)
|
87 |
+
html = '<div><p id="hldoc" style="font-size: 16px;">' + html.strip(' ') + '</p></div>'
|
88 |
+
|
89 |
+
return latency, ci, ngram_len, html
|
90 |
+
|
91 |
+
with gr.Blocks() as demo:
|
92 |
+
with gr.Column():
|
93 |
+
gr.HTML(
|
94 |
+
'''<h1 text-align="center">Creativity Index</h1>
|
95 |
+
|
96 |
+
<p style='font-size: 16px;'>Compute the <a href="">Creativity Index</a> of a piece of text.</p>
|
97 |
+
<p style='font-size: 16px;'>The computed Creativity Index is based on verbatim match and is supported by <a href="https://infini-gram.io">infini-gram</a>.</p>
|
98 |
+
'''
|
99 |
+
)
|
100 |
+
with gr.Row():
|
101 |
+
with gr.Column(scale=1, min_width=240):
|
102 |
+
index_desc = gr.Radio(choices=INDEX_DESCS, label='Corpus', value=INDEX_DESCS[0])
|
103 |
+
|
104 |
+
with gr.Column(scale=3):
|
105 |
+
creativity_query = gr.Textbox(placeholder='Enter a piece of text here', label='Query', interactive=True, lines=10)
|
106 |
+
with gr.Row():
|
107 |
+
creativity_clear = gr.ClearButton(value='Clear', variant='secondary', visible=True)
|
108 |
+
creativity_submit = gr.Button(value='Submit', variant='primary', visible=True)
|
109 |
+
creativity_latency = gr.Textbox(label='Latency (milliseconds)', interactive=False, lines=1)
|
110 |
+
|
111 |
+
with gr.Column(scale=4):
|
112 |
+
creativity_ci = gr.Label(value='', label='Creativity Index')
|
113 |
+
creativity_ngram_len = gr.Slider(minimum=NGRAM_LEN_MIN, maximum=NGRAM_LEN_MAX, value=NGRAM_LEN_DEFAULT, step=1, label='Length of n-gram')
|
114 |
+
creativity_html = gr.HTML(value='', label='Coverage')
|
115 |
+
|
116 |
+
creativity_clear.add([creativity_query, creativity_latency, creativity_ci, creativity_html])
|
117 |
+
creativity_submit.click(creativity, inputs=[index_desc, creativity_query], outputs=[creativity_latency, creativity_ci, creativity_ngram_len, creativity_html], api_name=False)
|
118 |
+
|
119 |
+
demo.queue(
|
120 |
+
default_concurrency_limit=DEFAULT_CONCURRENCY_LIMIT,
|
121 |
+
max_size=MAX_SIZE,
|
122 |
+
api_open=False,
|
123 |
+
).launch(
|
124 |
+
max_threads=MAX_THREADS,
|
125 |
+
debug=DEBUG,
|
126 |
+
show_api=False,
|
127 |
+
)
|
constants.py
ADDED
@@ -0,0 +1,24 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
|
3 |
+
# options
|
4 |
+
INDEX_BY_DESC = {
|
5 |
+
'Dolma-v1.7 (2.6T tokens)': 'v4_dolma-v1_7_llama',
|
6 |
+
'RedPajama (1.4T tokens)': 'v4_rpj_llama_s4',
|
7 |
+
'Pile-train (380B tokens)': 'v4_piletrain_llama',
|
8 |
+
'C4-train (200B tokens)': 'v4_c4train_llama',
|
9 |
+
'Pile-val (390M tokens)': 'v4_pileval_llama',
|
10 |
+
}
|
11 |
+
INDEX_DESCS = list(INDEX_BY_DESC.keys())
|
12 |
+
|
13 |
+
# API limits and defaults
|
14 |
+
MAX_QUERY_CHARS = int(os.environ.get('MAX_QUERY_CHARS', 1000))
|
15 |
+
NGRAM_LEN_DEFAULT = int(os.environ.get('NGRAM_LEN_DEFAULT', 8))
|
16 |
+
NGRAM_LEN_MIN = int(os.environ.get('NGRAM_LEN_MIN', 5))
|
17 |
+
NGRAM_LEN_MAX = int(os.environ.get('NGRAM_LEN_MAX', 11))
|
18 |
+
|
19 |
+
# HF demo
|
20 |
+
API_URL = os.environ.get('API_URL', None)
|
21 |
+
DEFAULT_CONCURRENCY_LIMIT = os.environ.get('DEFAULT_CONCURRENCY_LIMIT', 10)
|
22 |
+
MAX_SIZE = os.environ.get('MAX_SIZE', 100)
|
23 |
+
MAX_THREADS = os.environ.get('MAX_THREADS', 40)
|
24 |
+
DEBUG = (os.environ.get('DEBUG', 'False') != 'False')
|
requirements.txt
ADDED
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
torch==1.13.1
|
2 |
+
transformers==4.31.0
|
3 |
+
tokenizers==0.13.3
|
4 |
+
sentencepiece==0.1.96
|
5 |
+
huggingface_hub==0.14.1
|
6 |
+
requests
|