Update app.py

app.py

@@ -1,17 +1,21 @@
 import gradio as gr
-from transformers import pipeline, BartForConditionalGeneration, BartTokenizer
+from transformers import pipeline, AutoModelForSeq2SeqLM, AutoTokenizer
 import torch
 
 # 1. Configuration
 MODEL_ID = "VoltIC/Automated-Text-Summarizer"
 SUBFOLDER = "summarizer_model"
 
-print("🚀 Phase 1: Loading BART-specific Tokenizer...")
-# We use BartTokenizer directly to bypass the config.json check
-tokenizer = BartTokenizer.from_pretrained(MODEL_ID, subfolder=SUBFOLDER)
+print("🚀 Phase 1: Loading Tokenizer...")
+# use_fast=False is the magic fix for the NoneType/vocab_file error
+tokenizer = AutoTokenizer.from_pretrained(
+    MODEL_ID, 
+    subfolder=SUBFOLDER, 
+    use_fast=False
+)
 
-print("🚀 Phase 2: Loading BART Weights...")
-model = BartForConditionalGeneration.from_pretrained(MODEL_ID, subfolder=SUBFOLDER)
+print("🚀 Phase 2: Loading Model Weights...")
+model = AutoModelForSeq2SeqLM.from_pretrained(MODEL_ID, subfolder=SUBFOLDER)
 
 print("🚀 Phase 3: Building Pipeline...")
 summarizer = pipeline(
@@ -24,9 +28,17 @@ summarizer = pipeline(
 print("✅ SUCCESS: Summarizer Ready!")
 
 def summarize_text(text):
-    if not text: return "Input empty."
+    if not text: return "Input is empty."
+    # Truncation=True ensures we don't exceed the 1024 token limit
     results = summarizer(text, max_length=100, min_length=30, truncation=True)
     return results[0]['summary_text']
 
-demo = gr.Interface(fn=summarize_text, inputs="textbox", outputs="textbox")
-demo.launch()
+demo = gr.Interface(
+    fn=summarize_text, 
+    inputs=gr.Textbox(lines=10, placeholder="Paste text here..."), 
+    outputs="textbox",
+    title="AI Text Summarizer"
+)
+
+if __name__ == "__main__":
+    demo.launch()