Asanaly
commited on
Create summarizer.py
Browse files- summarizer.py +22 -0
summarizer.py
ADDED
|
@@ -0,0 +1,22 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
from transformers import BartTokenizer, BartForConditionalGeneration
|
| 2 |
+
|
| 3 |
+
tokenizer = BartTokenizer.from_pretrained("facebook/bart-large-cnn")
|
| 4 |
+
model = BartForConditionalGeneration.from_pretrained("facebook/bart-large-cnn")
|
| 5 |
+
|
| 6 |
+
def generate_summary(text: str) -> str:
|
| 7 |
+
text = text.strip()
|
| 8 |
+
|
| 9 |
+
if len(text) < 50:
|
| 10 |
+
return "Text is too short to summarize."
|
| 11 |
+
|
| 12 |
+
inputs = tokenizer([text], max_length=1024, truncation=True, return_tensors="pt")
|
| 13 |
+
|
| 14 |
+
summary_ids = model.generate(
|
| 15 |
+
inputs["input_ids"],
|
| 16 |
+
num_beams=4,
|
| 17 |
+
max_length=200,
|
| 18 |
+
min_length=40,
|
| 19 |
+
early_stopping=True
|
| 20 |
+
)
|
| 21 |
+
|
| 22 |
+
return tokenizer.decode(summary_ids[0], skip_special_tokens=True)
|