add t5 abstractive summarizer

app.py

@@ -1,24 +1,53 @@
+import torch
 import streamlit as st
+from extractive_summarizer.model_processors import Summarizer
+from transformers import T5Tokenizer, T5ForConditionalGeneration, T5Config
 
+def abstractive_summarizer(text : str):
+
+    model = T5ForConditionalGeneration.from_pretrained('t5-large')
+    tokenizer = T5Tokenizer.from_pretrained('t5-large')
+    device = torch.device('cpu')
+
+    preprocess_text = text.strip().replace("\n", "")
+    t5_prepared_text = "summarize: " + preprocess_text
+    tokenized_text = tokenizer.encode(t5_prepared_text, return_tensors="pt").to(device)
+
+    # summmarize 
+    summary_ids = model.generate(tokenized_text,
+                                        num_beams=4,
+                                        no_repeat_ngram_size=2,
+                                        min_length=30,
+                                        max_length=100,
+                                        early_stopping=True)
+    abs_summarized_text = tokenizer.decode(summary_ids[0], skip_special_tokens=True)
+
+    return abs_summarized_text
 
 if __name__ == "__main__":
-    # adding modules to path
-    import sys
-    sys.path.append("../extractive_summarizer")
+
     st.title("Text Summarizer 📝")
     summarize_type = st.sidebar.selectbox("Summarization type", options=["Extractive", "Abstractive"])
 
     inp_text = st.text_input("Enter the text here")
 
-    if summarize_type == "Extractive":
-        from extractive_summarizer.model_processors import Summarizer
+    # view summarized text (expander)
+    with st.expander("View input text"):
+        st.write(inp_text)
+
+    summarize = st.button("Summarize")
 
-        # init model
-        model = Summarizer()
-        summarize = st.button("Summarize")
-        if summarize:
-            with st.expander("View input text"):
-                st.write(inp_text)
-            st.subheader("Summarized text")
+    # called on toggle button [summarize]
+    if summarize:
+        if summarize_type == "Extractive":
+            # extractive summarizer
+            # init model
+            model = Summarizer()
             summarized_text = model(inp_text, num_sentences=5)
-            st.info(summarized_text)
+      
+        elif summarize_type == "Abstractive":
+            summarized_text = abstractive_summarizer(inp_text)
+
+        # final summarized output    
+        st.subheader("Summarized text")
+        st.info(summarized_text)