Create app.py

app.py

@@ -0,0 +1,79 @@
+# Import necessary libraries
+import PyPDF2
+from transformers import RagTokenizer, RagRetriever, RagSequenceForGeneration
+from sentence_transformers import SentenceTransformer
+import torch
+from google.colab import files
+
+# Step 1: Upload the PDF file
+print("Please upload a PDF file containing the chapter.")
+uploaded = files.upload()
+
+# Extract the name of the uploaded file
+file_name = list(uploaded.keys())[0]
+
+# Step 2: Extract text from the PDF
+def extract_text_from_pdf(file_path):
+    text = ""
+    with open(file_path, 'rb') as file:
+        pdf_reader = PyPDF2.PdfReader(file)
+        for page in pdf_reader.pages:
+            text += page.extract_text()
+    return text
+
+chapter_text = extract_text_from_pdf(file_name)
+print("Text extracted from the PDF successfully!")
+
+# Step 3: Split the text into smaller passages
+def split_text_into_chunks(text, chunk_size=500):
+    """Split the text into chunks of size chunk_size."""
+    words = text.split()
+    chunks = []
+    for i in range(0, len(words), chunk_size):
+        chunk = " ".join(words[i:i + chunk_size])
+        chunks.append(chunk)
+    return chunks
+
+passages = split_text_into_chunks(chapter_text)
+print(f"Chapter split into {len(passages)} passages for RAG processing.")
+
+# Step 4: Initialize the RAG model and tokenizer
+device = "cuda" if torch.cuda.is_available() else "cpu"
+
+# Load the RAG model, tokenizer, and retriever
+tokenizer = RagTokenizer.from_pretrained("facebook/rag-token-nq")
+retriever = RagRetriever.from_pretrained(
+    "facebook/rag-token-nq",
+    index_name="custom",
+    passages=passages,  # Set passages as the custom index
+    use_dummy_dataset=True,  # Dummy dataset required for custom index
+)
+model = RagSequenceForGeneration.from_pretrained("facebook/rag-token-nq").to(device)
+
+# Step 5: Encode passages into embeddings for retrieval
+sentence_model = SentenceTransformer("sentence-transformers/all-mpnet-base-v2")
+passage_embeddings = sentence_model.encode(passages, convert_to_tensor=True)
+retriever.index.set_passages(passages, passage_embeddings.cpu().detach().numpy())
+print("Passages indexed successfully!")
+
+# Step 6: Define a function to generate answers
+def generate_answer(question, passages):
+    inputs = tokenizer.prepare_seq2seq_batch(
+        questions=[question], return_tensors="pt"
+    ).to(device)
+    generated_ids = model.generate(**inputs)
+    answer = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
+    return answer
+
+# Step 7: Interactive Question-Answering
+print("\nChapter is ready. You can now ask questions!")
+while True:
+    user_question = input("\nEnter your question (or type 'exit' to quit): ")
+    if user_question.lower() == "exit":
+        print("Exiting the application. Thank you!")
+        break
+    try:
+        answer = generate_answer(user_question, passages)
+        print(f"Answer: {answer}")
+    except Exception as e:
+        print(f"An error occurred: {e}")