Create app.py

app.py

@@ -0,0 +1,84 @@
+import pandas as pd
+import numpy as np
+import re
+import gradio as gr
+import nltk
+from nltk.tokenize import sent_tokenize
+nltk.download('punkt')
+
+def read_in_text(url):
+  with open(url, 'r') as file:
+    article = file.read()
+    return article
+    
+def clean_text(url):
+  text = read_in_text(url)
+  #converting the text to all lower case
+  text = text.lower()
+
+  #removing the dates, time and name of author
+  text = re.sub('(by[\s\w,|]+ - \d\d\/\d\d\/\d\d\s\d+:\d+\s\w{2}\s\w{2})|(by[\s\w|,]+\d\d,\s\d{4})', "", text)
+  return text
+  
+  
+from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
+from transformers import PegasusForConditionalGeneration, PegasusTokenizer
+
+import torch
+device = 'cuda' if torch.cuda.is_available() else 'cpu'
+print ("device ",device)
+
+
+model_1 = AutoModelForSeq2SeqLM.from_pretrained("ramsrigouthamg/t5-large-paraphraser-diverse-high-quality").to(device)
+tokenizer_1 = AutoTokenizer.from_pretrained("ramsrigouthamg/t5-large-paraphraser-diverse-high-quality")
+
+
+
+
+tokenizer_2 = PegasusTokenizer.from_pretrained('tuner007/pegasus_paraphrase')
+model_2 = PegasusForConditionalGeneration.from_pretrained('tuner007/pegasus_paraphrase').to(device)
+
+# Diverse Beam search
+def my_paraphrase(sentence, model, tokenizer):
+
+  text = "paraphrase: "+sentence + " </s>"
+  encoding = tokenizer.encode_plus(text,max_length =60, padding=True, return_tensors="pt", truncation=True)
+  input_ids,attention_mask  = encoding["input_ids"].to(device), encoding["attention_mask"].to(device)
+
+  model.eval()
+  diverse_beam_outputs = model.generate(
+    input_ids=input_ids,attention_mask=attention_mask,
+    max_length=60,
+    early_stopping=True,
+    num_beams=5,
+    num_beam_groups = 5,
+    num_return_sequences=5,
+    diversity_penalty = 0.70
+  )
+  sent = tokenizer.decode(diverse_beam_outputs[0], skip_special_tokens=True,clean_up_tokenization_spaces=True)
+  return sent
+  
+def return_output(file, models):
+
+  docs = file.name
+  sentence = clean_text(docs)
+
+  if models == 'T5':
+    model = model_1
+    tokenizer = tokenizer_1
+    
+  elif models == 'Pegasus':
+    model = model_2
+    tokenizer = tokenizer_2
+
+  output = " ".join([my_paraphrase(sent, model, tokenizer) for sent in sent_tokenize(sentence)])
+  new_output = output.replace('paraphrasedoutput:', "")
+  new_output = new_output.replace('.<n>', '.\n')
+  return new_output
+  
+demo = gr.Interface(return_output, inputs=[gr.inputs.File(label="File", optional=False),
+                                           gr.inputs.Dropdown(['Pegasus', 'T5'], type="value", default=None, label="Models", optional=False),],
+                                                 outputs=[gr.outputs.Textbox(label="Summary")])
+                                                 
+if __name__ == "__main__":
+    demo.launch(debug=True)