Spaces:
Runtime error
Runtime error
File size: 3,346 Bytes
82e9de9 374bb44 82e9de9 374bb44 47c546d 7180e60 e844261 47c546d 82e9de9 374bb44 1bca7c7 82e9de9 1bca7c7 a4d32fc 1bca7c7 a4d32fc 374bb44 a4d32fc 82e9de9 1bca7c7 47c546d a4d32fc e844261 1bca7c7 e844261 82e9de9 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 |
import gradio
import torch
from transformers import pipeline, AutoTokenizer, AutoModelForSeq2SeqLM
def shorten_text(text, min_length, max_length):
summarizer = pipeline("summarization", model="facebook/bart-large-cnn")
short_text = text[:1024]
summary = summarizer(short_text, max_length, min_length, do_sample=False)
print("** summary", summary)
return summary[0]["summary_text"]
def paraphrase_text(text, min_length, max_length):
tokenizer = AutoTokenizer.from_pretrained("Vamsi/T5_Paraphrase_Paws")
model = AutoModelForSeq2SeqLM.from_pretrained("Vamsi/T5_Paraphrase_Paws")
device = "cuda" if torch.cuda.is_available() else "cpu"
text_instruction = "paraphrase: " + text + " </s>"
encoding = tokenizer.encode_plus(text_instruction, padding="longest", return_tensors="pt")
input_ids, attention_masks = encoding["input_ids"].to(device), encoding["attention_mask"].to(device)
outputs = model.generate(
input_ids=input_ids, attention_mask=attention_masks,
max_length=max_length,
do_sample=True,
top_k=120,
top_p=0.95,
early_stopping=True,
num_return_sequences=5
)
line = tokenizer.decode(outputs[0], skip_special_tokens=True, clean_up_tokenization_spaces=True)
print("** outputs", len(outputs), line)
return line
def modify_text(mode, text, min_length, max_length):
if mode == "shorten":
return shorten_text(text, min_length, max_length)
else:
return paraphrase_text(text, min_length, max_length)
gradio_interface = gradio.Interface(
fn=modify_text,
inputs=[
gradio.Radio(["shorten", "paraphrase"], label="Mode"),
"text",
gradio.Slider(5, 200, value=30, label="Min length"),
gradio.Slider(5, 500, value=130, label="Max length")
],
outputs="text",
examples=[
["shorten",
"""A beautiful golden sun is setting. The sky is on fire. A large neon sign rises into shot. It rests on top of a skyscraper and fills the frame. The building is neither past nor future in design but a bit of both.
Slowly we pan downwards revealing the city that spreads below. A glittering conglomeration of elevated transport tubes, smaller square buildings which are merely huge, with, here and there, the comparatively minuscule relics of previous ages of architecture, pavement level awnings suggesting restaurants and shops. Transparent tubes carry whizzing transport cages past us. An elevated highway carrying traffic composed primarily of large transport lorries passes through frame.
As we descend, the sunlight is blocked out and street lights & neon signs take over as illumination. Eventually we reach the upper levels of a plush shopping precinct.
Xmas decorations are everywhere. People are busy buying, ogling, discussing, choosing wisely from the goodies on display. Shoppers are going by laden with superbly packaged goods. The shop windows are full of elaborately boxed and be-ribboned who-knows-what. In one window is a bank of TV sets on the great majority of the screens is the face of Mr. Helpmann the Deputy Minister of Information. He is being interviewed. No-one bothers to listen to Helpmann.""",
30, 130]
],
title="Text shortener/paraphraser",
description="Shortening texts using `facebook/bart-large-cnn`, paraphrasing texts using `Vamsi/T5_Paraphrase_Paws`.",
article="© Tom Söderlund 2022"
)
gradio_interface.launch()
|