Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
@@ -1,3 +1,5 @@
|
|
|
|
|
|
1 |
import gradio as gr
|
2 |
from PyPDF4 import PdfFileReader
|
3 |
import tiktoken
|
@@ -10,71 +12,88 @@ def extract_text_from_pdf(file_path):
|
|
10 |
text += pdf.getPage(page_num).extractText()
|
11 |
return text
|
12 |
|
13 |
-
def tokenize(text,model="gpt-3.5-turbo"):
|
14 |
tokenizer = tiktoken.encoding_for_model(model)
|
15 |
-
tokens = tokenizer.encode(
|
16 |
-
text,
|
17 |
-
disallowed_special=()
|
18 |
-
)
|
19 |
return tokens
|
20 |
|
21 |
def count_tokens(text):
|
22 |
return len(tokenize(text))
|
23 |
|
24 |
-
def
|
25 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
26 |
paper_text = extract_text_from_pdf(file.name)
|
27 |
-
return
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
28 |
|
29 |
def chunk_text(text, max_char, overlap):
|
30 |
chunks = []
|
31 |
start = 0
|
32 |
end = max_char
|
33 |
-
print(f"max char: {max_char}")
|
34 |
while start < len(text):
|
35 |
if end >= len(text):
|
36 |
end = len(text)
|
37 |
chunk = text[start:end]
|
38 |
-
|
39 |
-
chunks.append(chunk)
|
40 |
start += max_char - overlap
|
41 |
end = start + max_char
|
42 |
return chunks
|
43 |
|
44 |
-
def chunk_file(file, max_char,overlap):
|
45 |
-
# Extract text from the PDF file
|
46 |
text = extract_text_from_pdf(file.name)
|
47 |
chunks = chunk_text(text, max_char, overlap)
|
|
|
|
|
|
|
48 |
|
49 |
-
|
|
|
|
|
|
|
|
|
50 |
|
51 |
with gr.Blocks() as demo:
|
52 |
-
gr.
|
53 |
-
|
54 |
-
|
55 |
-
|
56 |
-
|
57 |
-
|
58 |
-
|
|
|
59 |
|
60 |
-
|
61 |
-
|
62 |
-
|
63 |
-
btn_chunk.click(chunk_file,inputs=[docs_input,sl_max_char_per_chunk,sl_overlap],outputs=[tb_chunked_text])
|
64 |
-
with gr.Tab("Text"):
|
65 |
-
text_input = gr.Textbox(label='Insert your text here')
|
66 |
-
text_tb_tokenCount = gr.Textbox(label='Number of tokens')
|
67 |
-
text_input.change(count_tokens,inputs=[text_input],outputs=[text_tb_tokenCount])
|
68 |
-
text_sl_max_char_per_chunk = gr.Slider(1000, 30000, value=2000, label="Number of characters", info="Choose a number of characters per chunk")
|
69 |
-
text_sl_overlap = gr.Slider(0, 20000, value=400, label="Overlap", info="Choose overlap size")
|
70 |
|
71 |
-
|
72 |
-
text_tb_chunked_text = gr.Textbox(label='Result')
|
73 |
-
def format_chunks(text,max_char,overlap):
|
74 |
-
return '\n\n[xxxxxxxxxxxxxxxx]\n\n'.join(chunk_text(text,max_char,overlap))
|
75 |
-
text_btn_chunk.click(format_chunks,
|
76 |
-
inputs=[text_input,text_sl_max_char_per_chunk,text_sl_overlap],
|
77 |
-
outputs=[text_tb_chunked_text])
|
78 |
|
79 |
-
|
80 |
-
demo.launch(debug=True,share=False)
|
|
|
1 |
+
import os
|
2 |
+
import zipfile
|
3 |
import gradio as gr
|
4 |
from PyPDF4 import PdfFileReader
|
5 |
import tiktoken
|
|
|
12 |
text += pdf.getPage(page_num).extractText()
|
13 |
return text
|
14 |
|
15 |
+
def tokenize(text, model="gpt-3.5-turbo"):
|
16 |
tokenizer = tiktoken.encoding_for_model(model)
|
17 |
+
tokens = tokenizer.encode(text, disallowed_special=())
|
|
|
|
|
|
|
18 |
return tokens
|
19 |
|
20 |
def count_tokens(text):
|
21 |
return len(tokenize(text))
|
22 |
|
23 |
+
def analyse_text(text):
|
24 |
+
num_tokens = count_tokens(text)
|
25 |
+
result = []
|
26 |
+
try:
|
27 |
+
result.append(f"Text length: {len(text)}")
|
28 |
+
result.append(f"Token counts: {num_tokens}")
|
29 |
+
result.append(f"Char per token: {'%.1f' % (len(text)/num_tokens)}")
|
30 |
+
except:
|
31 |
+
result = 'no text'
|
32 |
+
return '\n'.join(result)
|
33 |
+
|
34 |
+
def analyse_file(file):
|
35 |
paper_text = extract_text_from_pdf(file.name)
|
36 |
+
return paper_text
|
37 |
+
|
38 |
+
def write_chunks_to_files(chunks):
|
39 |
+
file_paths = []
|
40 |
+
for i, chunk in enumerate(chunks, start=1):
|
41 |
+
file_path = f"chunk_{i}.txt"
|
42 |
+
with open(file_path, "w") as file:
|
43 |
+
file.write(chunk)
|
44 |
+
file_paths.append(file_path)
|
45 |
+
return file_paths
|
46 |
+
|
47 |
+
def write_chunks_to_zip(chunks):
|
48 |
+
file_paths = write_chunks_to_files(chunks)
|
49 |
+
zip_file_name = "chunks.zip"
|
50 |
+
with zipfile.ZipFile(zip_file_name, 'w') as zipf:
|
51 |
+
for file in file_paths:
|
52 |
+
zipf.write(file)
|
53 |
+
os.remove(file) # Remove the file after writing it into the zip
|
54 |
+
return zip_file_name
|
55 |
|
56 |
def chunk_text(text, max_char, overlap):
|
57 |
chunks = []
|
58 |
start = 0
|
59 |
end = max_char
|
|
|
60 |
while start < len(text):
|
61 |
if end >= len(text):
|
62 |
end = len(text)
|
63 |
chunk = text[start:end]
|
64 |
+
num_tokens = count_tokens(chunk)
|
65 |
+
chunks.append((chunk, len(chunk), num_tokens))
|
66 |
start += max_char - overlap
|
67 |
end = start + max_char
|
68 |
return chunks
|
69 |
|
70 |
+
def chunk_file(file, max_char, overlap):
|
|
|
71 |
text = extract_text_from_pdf(file.name)
|
72 |
chunks = chunk_text(text, max_char, overlap)
|
73 |
+
formatted_chunks = [f"Chunk[{i}]: Size: {len(c[0])} chars, {c[2]} tokens" for i, c in enumerate(chunks, start=1)]
|
74 |
+
zip_file_path = write_chunks_to_zip([c[0] for c in chunks])
|
75 |
+
return '\n'.join(formatted_chunks), zip_file_path
|
76 |
|
77 |
+
def chunk_and_zip_text(text, max_char, overlap):
|
78 |
+
chunks = chunk_text(text, max_char, overlap)
|
79 |
+
formatted_chunks = [f"Chunk[{i}]: Size: {len(c[0])} chars, {c[2]} tokens" for i, c in enumerate(chunks, start=1)]
|
80 |
+
zip_file_path = write_chunks_to_zip([c[0] for c in chunks])
|
81 |
+
return '\n'.join(formatted_chunks), zip_file_path
|
82 |
|
83 |
with gr.Blocks() as demo:
|
84 |
+
docs_input = gr.File(file_count="single", file_types=[".pdf"])
|
85 |
+
text_to_chunk = gr.Textbox(label='Text to chunk',show_copy_button=True)
|
86 |
+
tb_analysis = gr.Textbox(label='Text Analysis')
|
87 |
+
sl_max_char_per_chunk = gr.Slider(1000, 300000, value=10000, label="Number of characters", info="Choose a number of characters per chunk")
|
88 |
+
sl_overlap = gr.Slider(0, 20000, value=400, label="Overlap", info="Choose overlap size")
|
89 |
+
btn_chunk = gr.Button("Chunk text")
|
90 |
+
tb_chunked_text = gr.Textbox(label='Chunks Info')
|
91 |
+
download_link = gr.File(label='Download Chunks')
|
92 |
|
93 |
+
# Call analyse_file when a file is uploaded and display the results in tb_analysis
|
94 |
+
docs_input.upload(analyse_file,inputs=[docs_input], outputs=[text_to_chunk])
|
95 |
+
text_to_chunk.change(analyse_text,inputs=[text_to_chunk],outputs=[tb_analysis])
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
96 |
|
97 |
+
btn_chunk.click(chunk_and_zip_text, inputs=[text_to_chunk, sl_max_char_per_chunk, sl_overlap], outputs=[tb_chunked_text, download_link])
|
|
|
|
|
|
|
|
|
|
|
|
|
98 |
|
99 |
+
demo.launch(debug=True, share=False)
|
|