Update app.py

app.py

@@ -1,36 +1,28 @@
 import gradio as gr
-from ctransformers import AutoModelForCausalLM
+from transformers import T5Tokenizer, T5ForConditionalGeneration
 
-# Load GGUF model directly from Hugging Face (no need to pre-download)
-model = AutoModelForCausalLM.from_pretrained(
-    "TheBloke/Mistral-7B-Instruct-v0.1-GGUF",  # repo_id
-    model_file="mistral-7b-instruct-v0.1.Q4_K_M.gguf",  # exact filename
-    model_type="mistral",  # required
-    max_new_tokens=2048,
-    temperature=0.9,
-    repetition_penalty=1.1,
-    top_p=0.95
-)
-
-# Define the Q&A generation function
-def generate_qa(text):
-    prompt = f"""### Instruction:
-Based on the following SAP Note, generate exactly 20 unique and informative question-answer pairs.
-Each question must refer to the SAP note number from text if additional context is needed.
+# Load tokenizer and model (small T5 variant, CPU only)
+model_name = "t5-small"
 
-### Input:
-{text}
+tokenizer = T5Tokenizer.from_pretrained(model_name)
+model = T5ForConditionalGeneration.from_pretrained(model_name)
 
-### Response:"""
-    return model(prompt).strip()
+def generate_text(input_text):
+    # Prepare input tokens
+    input_ids = tokenizer.encode(input_text, return_tensors="pt")
+    # Generate output tokens (max length 100)
+    outputs = model.generate(input_ids, max_length=100, num_beams=5, early_stopping=True)
+    # Decode to text
+    result = tokenizer.decode(outputs[0], skip_special_tokens=True)
+    return result
 
-# Create Gradio UI
+# Build Gradio interface
 demo = gr.Interface(
-    fn=generate_qa,
-    inputs=gr.Textbox(lines=20, label="SAP Note Text"),
-    outputs=gr.Textbox(lines=30, label="Generated Q&A Pairs"),
-    title="SAP Note Q&A Generator (Mistral GGUF - CPU)",
-    description="Paste SAP Note content to generate 20 Q&A pairs using Mistral 7B (Quantized GGUF model)"
+    fn=generate_text,
+    inputs=gr.Textbox(lines=5, label="Input Text"),
+    outputs=gr.Textbox(lines=5, label="Generated Text"),
+    title="Simple T5 Text Generator (CPU)",
+    description="Enter some text, T5 will generate a continuation or answer."
 )
 
 if __name__ == "__main__":