⚡ Switched to Flan-T5 for faster response on CPU

app.py

@@ -6,6 +6,7 @@ import torch
 import gradio as gr
 from PyPDF2 import PdfReader
 from docx import Document as DocxDocument
+from transformers import pipeline, AutoTokenizer, AutoModelForSeq2SeqLM
 from pptx import Presentation
 from transformers import pipeline, AutoTokenizer, AutoModelForCausalLM
 from langchain_community.vectorstores import FAISS
@@ -13,16 +14,19 @@ from langchain_community.embeddings import HuggingFaceEmbeddings
 from langchain.text_splitter import RecursiveCharacterTextSplitter
 from langchain_core.documents import Document
 
-# Load Reasoning Model
-model_id = "microsoft/phi-3-mini-128k-instruct"
+# Load Reasoning Model (lightweight + CPU friendly)
+model_id = "google/flan-t5-base"
 tokenizer = AutoTokenizer.from_pretrained(model_id)
-model = AutoModelForCausalLM.from_pretrained(
-    model_id,
-    torch_dtype=torch.bfloat16,
-    device_map="auto"
+model = AutoModelForSeq2SeqLM.from_pretrained(model_id)
+
+reasoning_pipeline = pipeline(
+    "text2text-generation",
+    model=model,
+    tokenizer=tokenizer,
+    max_new_tokens=512,
+    temperature=0.7,
+    top_p=0.9
 )
-reasoning_pipeline = pipeline("text-generation", model=model, tokenizer=tokenizer,
-                              max_new_tokens=512, temperature=0.7, top_p=0.9)
 
 # Embedding Model
 embedding_model = HuggingFaceEmbeddings(model_name="all-MiniLM-L6-v2")