Spaces:
Running
Running
Pclanglais
commited on
Commit
•
750020e
1
Parent(s):
100e33a
Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,146 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import transformers
|
2 |
+
import re
|
3 |
+
from transformers import AutoConfig, AutoTokenizer, AutoModel, AutoModelForCausalLM
|
4 |
+
from vllm import LLM, SamplingParams
|
5 |
+
import torch
|
6 |
+
import gradio as gr
|
7 |
+
import json
|
8 |
+
import os
|
9 |
+
import shutil
|
10 |
+
import requests
|
11 |
+
import chromadb
|
12 |
+
import difflib
|
13 |
+
import pandas as pd
|
14 |
+
from chromadb.config import Settings
|
15 |
+
from chromadb.utils import embedding_functions
|
16 |
+
|
17 |
+
# Define the device
|
18 |
+
device = "cuda" if torch.cuda.is_available() else "cpu"
|
19 |
+
|
20 |
+
model_checkpoint = "PleIAs/Estienne"
|
21 |
+
token_classifier = pipeline(
|
22 |
+
"token-classification", model=editorial_model, aggregation_strategy="simple", device=device
|
23 |
+
)
|
24 |
+
|
25 |
+
tokenizer = AutoTokenizer.from_pretrained(editorial_model, model_max_length=512)
|
26 |
+
|
27 |
+
|
28 |
+
def split_text(text, max_tokens=500):
|
29 |
+
# Split the text by newline characters
|
30 |
+
parts = text.split("\n")
|
31 |
+
chunks = []
|
32 |
+
current_chunk = ""
|
33 |
+
|
34 |
+
for part in parts:
|
35 |
+
# Add part to current chunk
|
36 |
+
if current_chunk:
|
37 |
+
temp_chunk = current_chunk + "\n" + part
|
38 |
+
else:
|
39 |
+
temp_chunk = part
|
40 |
+
|
41 |
+
# Tokenize the temporary chunk
|
42 |
+
num_tokens = len(tokenizer.tokenize(temp_chunk))
|
43 |
+
|
44 |
+
if num_tokens <= max_tokens:
|
45 |
+
current_chunk = temp_chunk
|
46 |
+
else:
|
47 |
+
if current_chunk:
|
48 |
+
chunks.append(current_chunk)
|
49 |
+
current_chunk = part
|
50 |
+
|
51 |
+
if current_chunk:
|
52 |
+
chunks.append(current_chunk)
|
53 |
+
|
54 |
+
# If no newlines were found and still exceeding max_tokens, split further
|
55 |
+
if len(chunks) == 1 and len(tokenizer.tokenize(chunks[0])) > max_tokens:
|
56 |
+
long_text = chunks[0]
|
57 |
+
chunks = []
|
58 |
+
while len(tokenizer.tokenize(long_text)) > max_tokens:
|
59 |
+
split_point = len(long_text) // 2
|
60 |
+
while split_point < len(long_text) and not re.match(r'\s', long_text[split_point]):
|
61 |
+
split_point += 1
|
62 |
+
# Ensure split_point does not go out of range
|
63 |
+
if split_point >= len(long_text):
|
64 |
+
split_point = len(long_text) - 1
|
65 |
+
chunks.append(long_text[:split_point].strip())
|
66 |
+
long_text = long_text[split_point:].strip()
|
67 |
+
if long_text:
|
68 |
+
chunks.append(long_text)
|
69 |
+
|
70 |
+
return chunks
|
71 |
+
|
72 |
+
|
73 |
+
#Curtesy of claude
|
74 |
+
def generate_html_diff(old_text, new_text):
|
75 |
+
d = difflib.Differ()
|
76 |
+
diff = list(d.compare(old_text.split(), new_text.split()))
|
77 |
+
|
78 |
+
html_diff = []
|
79 |
+
for word in diff:
|
80 |
+
if word.startswith(' '):
|
81 |
+
html_diff.append(word[2:])
|
82 |
+
elif word.startswith('+ '):
|
83 |
+
html_diff.append(f'<span style="background-color: #90EE90;">{word[2:]}</span>')
|
84 |
+
# We're not adding anything for words that start with '- '
|
85 |
+
|
86 |
+
return ' '.join(html_diff)
|
87 |
+
|
88 |
+
# Class to encapsulate the Falcon chatbot
|
89 |
+
class MistralChatBot:
|
90 |
+
def __init__(self, system_prompt="Le dialogue suivant est une conversation"):
|
91 |
+
self.system_prompt = system_prompt
|
92 |
+
|
93 |
+
def predict(self, user_message):
|
94 |
+
#We drop the newlines.
|
95 |
+
editorial_text = re.sub("\n", " ¶ ", user_message)
|
96 |
+
|
97 |
+
# Tokenize the prompt and check if it exceeds 500 tokens
|
98 |
+
num_tokens = len(tokenizer.tokenize(prompt))
|
99 |
+
|
100 |
+
if num_tokens > 500:
|
101 |
+
# Split the prompt into chunks
|
102 |
+
batch_prompts = split_text(prompt, max_tokens=500)
|
103 |
+
else:
|
104 |
+
batch_prompts = [prompt]
|
105 |
+
|
106 |
+
out = token_classifier(batch_prompts)
|
107 |
+
out = "".join(out)
|
108 |
+
generated_text = '<h2 style="text-align:center">Réponse</h3>\n<div class="generation">' + html_diff + "</div>"
|
109 |
+
return generated_text
|
110 |
+
|
111 |
+
# Create the Falcon chatbot instance
|
112 |
+
mistral_bot = MistralChatBot()
|
113 |
+
|
114 |
+
# Define the Gradio interface
|
115 |
+
title = "Éditorialisation"
|
116 |
+
description = "Un outil expérimental d'identification de la structure du texte à partir d'un encoder (Deberta)"
|
117 |
+
examples = [
|
118 |
+
[
|
119 |
+
"Qui peut bénéficier de l'AIP?", # user_message
|
120 |
+
0.7 # temperature
|
121 |
+
]
|
122 |
+
]
|
123 |
+
|
124 |
+
additional_inputs=[
|
125 |
+
gr.Slider(
|
126 |
+
label="Température",
|
127 |
+
value=0.2, # Default value
|
128 |
+
minimum=0.05,
|
129 |
+
maximum=1.0,
|
130 |
+
step=0.05,
|
131 |
+
interactive=True,
|
132 |
+
info="Des valeurs plus élevées donne plus de créativité, mais aussi d'étrangeté",
|
133 |
+
),
|
134 |
+
]
|
135 |
+
|
136 |
+
demo = gr.Blocks()
|
137 |
+
|
138 |
+
with gr.Blocks(theme='JohnSmith9982/small_and_pretty', css=css) as demo:
|
139 |
+
gr.HTML("""<h1 style="text-align:center">Correction d'OCR</h1>""")
|
140 |
+
text_input = gr.Textbox(label="Votre texte.", type="text", lines=1)
|
141 |
+
text_button = gr.Button("Identifier les structures éditoriales")
|
142 |
+
text_output = gr.HTML(label="Le texte corrigé")
|
143 |
+
text_button.click(mistral_bot.predict, inputs=text_input, outputs=[text_output])
|
144 |
+
|
145 |
+
if __name__ == "__main__":
|
146 |
+
demo.queue().launch()
|