Spaces:
Sleeping
Sleeping
Orion Weller
commited on
Commit
•
53fb3e2
1
Parent(s):
39ee16f
init
Browse files- app.py +242 -0
- chunked_data.jsonl +0 -0
- generated_data.json +0 -0
- packages.txt +1 -0
- requirements.txt +7 -0
app.py
ADDED
@@ -0,0 +1,242 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import streamlit as st
|
2 |
+
import os
|
3 |
+
import pathlib
|
4 |
+
import pandas as pd
|
5 |
+
from collections import defaultdict
|
6 |
+
import json
|
7 |
+
import ast
|
8 |
+
import copy
|
9 |
+
import re
|
10 |
+
import tqdm
|
11 |
+
|
12 |
+
|
13 |
+
import pandas as pd
|
14 |
+
from collections import Counter
|
15 |
+
import string
|
16 |
+
import os
|
17 |
+
import streamlit as st
|
18 |
+
import difflib
|
19 |
+
from html import escape
|
20 |
+
|
21 |
+
|
22 |
+
def generate_diff_html_word_level(text1, text2):
|
23 |
+
"""
|
24 |
+
Generates word-level difference between text1 and text2 as HTML, correctly handling spaces.
|
25 |
+
"""
|
26 |
+
# Splitting texts into words
|
27 |
+
words1 = text1.split()
|
28 |
+
words2 = text2.split()
|
29 |
+
|
30 |
+
diff = []
|
31 |
+
matcher = difflib.SequenceMatcher(None, words1, words2)
|
32 |
+
|
33 |
+
for opcode in matcher.get_opcodes():
|
34 |
+
tag, i1, i2, j1, j2 = opcode
|
35 |
+
if tag == 'replace':
|
36 |
+
diff.append('<del style="background-color: #fbb6ce;">' + escape(' '.join(words1[i1:i2])) + '</del>')
|
37 |
+
diff.append('<ins style="background-color: #b7e4c7;">' + escape(' '.join(words2[j1:j2])) + '</ins>')
|
38 |
+
elif tag == 'delete':
|
39 |
+
diff.append('<del style="background-color: #fbb6ce;">' + escape(' '.join(words1[i1:i2])) + '</del>')
|
40 |
+
elif tag == 'insert':
|
41 |
+
diff.append('<ins style="background-color: #b7e4c7;">' + escape(' '.join(words2[j1:j2])) + '</ins>')
|
42 |
+
elif tag == 'equal':
|
43 |
+
diff.append(escape(' '.join(words1[i1:i2])))
|
44 |
+
|
45 |
+
# Construct final HTML string
|
46 |
+
final_html = ' '.join(diff).replace('</del> <ins', '</del> <ins')
|
47 |
+
return f'<pre style="white-space: pre-wrap;">{final_html}</pre>'
|
48 |
+
|
49 |
+
|
50 |
+
|
51 |
+
os.environ["PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION"] = "python"
|
52 |
+
st.set_page_config(layout="wide")
|
53 |
+
|
54 |
+
current_checkboxes = []
|
55 |
+
query_input = None
|
56 |
+
|
57 |
+
@st.cache_data
|
58 |
+
def convert_df(df):
|
59 |
+
# IMPORTANT: Cache the conversion to prevent computation on every rerun
|
60 |
+
return df.to_csv(path_or_buf=None, index=False, quotechar='"').encode('utf-8')
|
61 |
+
|
62 |
+
|
63 |
+
|
64 |
+
@st.cache_data
|
65 |
+
def escape_markdown(text):
|
66 |
+
# List of characters to escape
|
67 |
+
# Adding backslash to the list of special characters to escape itself as well
|
68 |
+
text = text.replace("``", "\"")
|
69 |
+
text = text.replace("$", "\$")
|
70 |
+
special_chars = ['\\', '`', '*', '_', '{', '}', '[', ']', '(', ')', '#', '+', '-', '.', '!', '|', "$"]
|
71 |
+
|
72 |
+
# Escaping each special character
|
73 |
+
escaped_text = "".join(f"\\{char}" if char in special_chars else char for char in text)
|
74 |
+
|
75 |
+
return escaped_text
|
76 |
+
|
77 |
+
|
78 |
+
|
79 |
+
|
80 |
+
if 'cur_instance_num' not in st.session_state:
|
81 |
+
st.session_state.cur_instance_num = -1
|
82 |
+
|
83 |
+
|
84 |
+
def validate(config_option, file_loaded):
|
85 |
+
if config_option != "None" and file_loaded is None:
|
86 |
+
st.error("Please upload a file for " + config_option)
|
87 |
+
st.stop()
|
88 |
+
|
89 |
+
|
90 |
+
|
91 |
+
|
92 |
+
|
93 |
+
with st.sidebar:
|
94 |
+
st.title("Options")
|
95 |
+
|
96 |
+
|
97 |
+
@st.cache_data
|
98 |
+
def load_chunked_data():
|
99 |
+
data = []
|
100 |
+
with open("chunked_data.jsonl", "r") as f:
|
101 |
+
for line in f:
|
102 |
+
data.append(json.loads(line))
|
103 |
+
# rename prompt column to text
|
104 |
+
df = pd.DataFrame(data)
|
105 |
+
df = df.rename(columns={"prompt": "text"})
|
106 |
+
return df
|
107 |
+
|
108 |
+
|
109 |
+
def load_generated_data():
|
110 |
+
with open("generated_data.json", "r") as fin:
|
111 |
+
data = json.load(fin)["outputs"]
|
112 |
+
|
113 |
+
new_insts = []
|
114 |
+
for key, value in data.items():
|
115 |
+
item = {
|
116 |
+
"venue": key
|
117 |
+
}
|
118 |
+
if type(value) == str:
|
119 |
+
value = ast.literal_eval(value)
|
120 |
+
|
121 |
+
if type(value) == dict:
|
122 |
+
for cur_key, cur_value in value.items():
|
123 |
+
item[cur_key] = cur_value
|
124 |
+
else:
|
125 |
+
raise ValueError(f"Invalid type {type(value)}: {value}")
|
126 |
+
new_insts.append(item)
|
127 |
+
|
128 |
+
return pd.DataFrame(new_insts)
|
129 |
+
|
130 |
+
original_df = load_chunked_data()
|
131 |
+
generated_data = load_generated_data()
|
132 |
+
|
133 |
+
|
134 |
+
def combine_text(item):
|
135 |
+
string_text = ""
|
136 |
+
for key, value in item.items():
|
137 |
+
if key == "venue" or value is None or value == "[]" or type(value) == float or len(value) == 0:
|
138 |
+
continue
|
139 |
+
string_text += f",{', '.join(value)}\n"
|
140 |
+
if "," == string_text[0]:
|
141 |
+
string_text = string_text[1:]
|
142 |
+
return string_text
|
143 |
+
|
144 |
+
|
145 |
+
original_map = {item["venue"]: item["text"] for item in original_df.to_dict(orient="records")}
|
146 |
+
generated_map = {item["venue"]: combine_text(item) for item in generated_data.to_dict(orient="records")}
|
147 |
+
|
148 |
+
|
149 |
+
col1, col2 = st.columns([1, 3], gap="large")
|
150 |
+
|
151 |
+
with st.sidebar:
|
152 |
+
st.success("All files uploaded")
|
153 |
+
|
154 |
+
with col1:
|
155 |
+
# breakpoint()
|
156 |
+
ids = original_df["venue"].tolist()
|
157 |
+
set_of_cols = set(ids)
|
158 |
+
container_for_nav = st.container()
|
159 |
+
name_of_columns = sorted([item for item in set_of_cols])
|
160 |
+
instances_to_use = name_of_columns
|
161 |
+
st.title("Instances")
|
162 |
+
|
163 |
+
def sync_from_drop():
|
164 |
+
if st.session_state.selectbox_instance == "Overview":
|
165 |
+
st.session_state.number_of_col = -1
|
166 |
+
st.session_state.cur_instance_num = -1
|
167 |
+
else:
|
168 |
+
index_of_obj = name_of_columns.index(st.session_state.selectbox_instance)
|
169 |
+
# print("Index of obj: ", index_of_obj, type(index_of_obj))
|
170 |
+
st.session_state.number_of_col = index_of_obj
|
171 |
+
st.session_state.cur_instance_num = index_of_obj
|
172 |
+
|
173 |
+
def sync_from_number():
|
174 |
+
st.session_state.cur_instance_num = st.session_state.number_of_col
|
175 |
+
# print("Session state number of col: ", st.session_state.number_of_col, type(st.session_state.number_of_col))
|
176 |
+
if st.session_state.number_of_col == -1:
|
177 |
+
st.session_state.selectbox_instance = "Overview"
|
178 |
+
else:
|
179 |
+
st.session_state.selectbox_instance = name_of_columns[st.session_state.number_of_col]
|
180 |
+
|
181 |
+
|
182 |
+
number_of_col = container_for_nav.number_input(min_value=-1, step=1, max_value=len(instances_to_use) - 1, on_change=sync_from_number, label=f"Select instance by index (up to **{len(instances_to_use) - 1}**)", key="number_of_col")
|
183 |
+
selectbox_instance = container_for_nav.selectbox("Select instance by ID", ["Overview"] + name_of_columns, on_change=sync_from_drop, key="selectbox_instance")
|
184 |
+
st.divider()
|
185 |
+
|
186 |
+
|
187 |
+
with col2:
|
188 |
+
# get instance number
|
189 |
+
inst_index = number_of_col
|
190 |
+
|
191 |
+
if inst_index >= 0:
|
192 |
+
inst_num = instances_to_use[inst_index]
|
193 |
+
|
194 |
+
st.markdown("<h1 style='text-align: center; color: black;text-decoration: underline;'>Editor</h1>", unsafe_allow_html=True)
|
195 |
+
|
196 |
+
|
197 |
+
container = st.container()
|
198 |
+
|
199 |
+
container.subheader(f"Venue: {inst_num}")
|
200 |
+
|
201 |
+
container.divider()
|
202 |
+
|
203 |
+
|
204 |
+
original_text = original_map[inst_num]
|
205 |
+
generated_text = generated_map[inst_num]
|
206 |
+
container.subheader(f"Original OCR Text")
|
207 |
+
original_input = container.markdown(original_text)
|
208 |
+
container.divider()
|
209 |
+
|
210 |
+
container.subheader(f"Generated Text")
|
211 |
+
generated_input = container.markdown(generated_text)
|
212 |
+
container.divider()
|
213 |
+
|
214 |
+
# print("Original text: ", original_text)
|
215 |
+
# print("Generated text: ", generated_text)
|
216 |
+
|
217 |
+
# Diff
|
218 |
+
if original_text is not None and generated_input is not None:
|
219 |
+
container.subheader("Diff")
|
220 |
+
processed_diff = generate_diff_html_word_level(original_map[inst_num], generated_map[inst_num])
|
221 |
+
with container.container(border=True):
|
222 |
+
st.markdown(processed_diff, unsafe_allow_html=True)
|
223 |
+
|
224 |
+
|
225 |
+
# editable text, starting from the generated text
|
226 |
+
editable_text = container.text_area("Edit the generated text", value=generated_text, height=300)
|
227 |
+
|
228 |
+
|
229 |
+
|
230 |
+
container.divider()
|
231 |
+
# download the editable text and venue name
|
232 |
+
st.download_button(
|
233 |
+
f"Download {inst_num} as CSV",
|
234 |
+
convert_df(pd.DataFrame([{"venue": inst_num, "text": editable_text}])),
|
235 |
+
f"{inst_num}.csv",
|
236 |
+
"text/csv",
|
237 |
+
key=f"download_{inst_num}"
|
238 |
+
)
|
239 |
+
|
240 |
+
# none checked
|
241 |
+
elif inst_index < 0:
|
242 |
+
st.title("Overview")
|
chunked_data.jsonl
ADDED
The diff for this file is too large to render.
See raw diff
|
|
generated_data.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|
packages.txt
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
default-jre
|
requirements.txt
ADDED
@@ -0,0 +1,7 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
pandas==2.0.3
|
2 |
+
streamlit==1.24.1
|
3 |
+
plotly==5.15.0
|
4 |
+
protobuf==3.20.0
|
5 |
+
beautifulsoup4==4.12.2
|
6 |
+
nltk==3.7
|
7 |
+
tqdm==4.66.2
|