Update app.py
Browse files
app.py
CHANGED
|
@@ -60,38 +60,76 @@ iface = gr.Interface(
|
|
| 60 |
|
| 61 |
iface.launch(share=True)
|
| 62 |
"""
|
|
|
|
| 63 |
import gradio as gr
|
| 64 |
import nltk
|
| 65 |
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
|
| 66 |
|
| 67 |
nltk.download('punkt')
|
| 68 |
|
|
|
|
| 69 |
def fragment_text(text, tokenizer):
|
| 70 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 71 |
|
| 72 |
def summarize_text(text, tokenizer, model):
|
| 73 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 74 |
|
| 75 |
checkpoint = "tclopess/bart_samsum"
|
| 76 |
tokenizer = AutoTokenizer.from_pretrained(checkpoint)
|
| 77 |
model = AutoModelForSeq2SeqLM.from_pretrained(checkpoint)
|
| 78 |
|
| 79 |
-
|
| 80 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 81 |
|
| 82 |
iface = gr.Interface(
|
| 83 |
-
fn=
|
| 84 |
-
inputs=
|
|
|
|
|
|
|
|
|
|
| 85 |
outputs=gr.Textbox(label="Summary:"),
|
|
|
|
| 86 |
title="Text Summarizer with Button",
|
| 87 |
)
|
| 88 |
|
| 89 |
-
|
| 90 |
-
iface.launch(share=True)
|
| 91 |
-
|
| 92 |
-
gr.Interface(
|
| 93 |
-
summarize_text_button,
|
| 94 |
-
inputs=None,
|
| 95 |
-
outputs=None,
|
| 96 |
-
title="Click to Summarize"
|
| 97 |
-
).launch()
|
|
|
|
| 60 |
|
| 61 |
iface.launch(share=True)
|
| 62 |
"""
|
| 63 |
+
|
| 64 |
import gradio as gr
|
| 65 |
import nltk
|
| 66 |
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
|
| 67 |
|
| 68 |
nltk.download('punkt')
|
| 69 |
|
| 70 |
+
|
| 71 |
def fragment_text(text, tokenizer):
|
| 72 |
+
sentences = nltk.tokenize.sent_tokenize(text)
|
| 73 |
+
max_len = tokenizer.max_len_single_sentence
|
| 74 |
+
|
| 75 |
+
chunks = []
|
| 76 |
+
chunk = ""
|
| 77 |
+
count = -1
|
| 78 |
+
|
| 79 |
+
for sentence in sentences:
|
| 80 |
+
count += 1
|
| 81 |
+
combined_length = len(tokenizer.tokenize(sentence)) + len(chunk)
|
| 82 |
+
|
| 83 |
+
if combined_length <= max_len:
|
| 84 |
+
chunk += sentence + " "
|
| 85 |
+
else:
|
| 86 |
+
chunks.append(chunk.strip())
|
| 87 |
+
chunk = sentence + " "
|
| 88 |
+
|
| 89 |
+
if chunk != "":
|
| 90 |
+
chunks.append(chunk.strip())
|
| 91 |
+
|
| 92 |
+
return chunks
|
| 93 |
+
|
| 94 |
|
| 95 |
def summarize_text(text, tokenizer, model):
|
| 96 |
+
chunks = fragment_text(text, tokenizer)
|
| 97 |
+
|
| 98 |
+
summaries = []
|
| 99 |
+
for chunk in chunks:
|
| 100 |
+
input = tokenizer(chunk, return_tensors='pt')
|
| 101 |
+
output = model.generate(**input)
|
| 102 |
+
summary = tokenizer.decode(*output, skip_special_tokens=True)
|
| 103 |
+
summaries.append(summary)
|
| 104 |
+
|
| 105 |
+
final_summary = " ".join(summaries)
|
| 106 |
+
return final_summary
|
| 107 |
+
|
| 108 |
|
| 109 |
checkpoint = "tclopess/bart_samsum"
|
| 110 |
tokenizer = AutoTokenizer.from_pretrained(checkpoint)
|
| 111 |
model = AutoModelForSeq2SeqLM.from_pretrained(checkpoint)
|
| 112 |
|
| 113 |
+
|
| 114 |
+
def summarize_and_display(text):
|
| 115 |
+
summary = summarize_text(text, tokenizer, model)
|
| 116 |
+
return summary
|
| 117 |
+
|
| 118 |
+
|
| 119 |
+
def start_summarization(text):
|
| 120 |
+
summary = summarize_and_display(text)
|
| 121 |
+
gr.update(summary)
|
| 122 |
+
|
| 123 |
|
| 124 |
iface = gr.Interface(
|
| 125 |
+
fn=start_summarization,
|
| 126 |
+
inputs=[
|
| 127 |
+
gr.Textbox(label="Enter text to summarize:"),
|
| 128 |
+
gr.Button(label="Summarize"),
|
| 129 |
+
],
|
| 130 |
outputs=gr.Textbox(label="Summary:"),
|
| 131 |
+
live=True,
|
| 132 |
title="Text Summarizer with Button",
|
| 133 |
)
|
| 134 |
|
| 135 |
+
iface.launch(share=True)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|