Create app.py

app.py

@@ -0,0 +1,36 @@
+import streamlit as st
+from flask import Flask, request, render_template, redirect
+from transformers import BartTokenizer, BartForConditionalGeneration
+
+# preprocess input
+# return input_ids matrix
+tokenizer = BartTokenizer.from_pretrained("sshleifer/distilbart-cnn-6-6")
+model = BartForConditionalGeneration.from_pretrained("sshleifer/distilbart-cnn-6-6")
+
+def preprocess(inp):
+    input_ids = tokenizer(inp, return_tensors="pt").input_ids
+    return input_ids
+def predict(input_ids):
+    outputs = model.generate(input_ids=input_ids)
+    res = tokenizer.batch_decode(outputs, skip_special_tokens=True)[0]
+    return res
+
+@app.route('/', methods=['POST', 'GET'])
+def index():
+    if request.method == 'POST':
+        inp = request.form['content']
+        inp_ids = preprocess(inp)
+        summary = predict(inp_ids)
+        return render_template('index.html', summary=summary)
+    else:
+        print("GETTING get")
+        return render_template('index.html', summary="Nothing to summarize")
+
+
+if __name__ == '__main__':
+    st.title("Text summary with fine-tuned Pegasus model")
+    with st.container():
+        txt = st.text_area('Text to analyze', ' ')
+        inp_ids = preprocess(txt)
+        st.write('Summary:', predict(inp_ids))
+