Update app.py

app.py

@@ -1,7 +1,7 @@
+import re
 from transformers import pipeline, AutoTokenizer, AutoModelForSequenceClassification
 from datasets import load_dataset
 
-# Model details
 MODEL_NAME = "Pisethan/sangapac-math"
 
 # Load model and tokenizer
@@ -15,34 +15,33 @@ except Exception as e:
 
 # Load dataset dynamically from Hugging Face or locally
 try:
-    dataset = load_dataset("Pisethan/sangapac-math-dataset")["train"]  # Load your dataset
-    dataset_dict = {entry["input"]: entry for entry in dataset}  # Create a dictionary for lookup
+    dataset = load_dataset("Pisethan/sangapac-math-dataset")["train"]
+    dataset_dict = {re.sub(r'\s+', ' ', entry["input"].strip()): entry for entry in dataset}
 except Exception as e:
     dataset_dict = {}
     print(f"Error loading dataset: {e}")
 
+def normalize_input(text):
+    return re.sub(r'\s+', ' ', text.strip())
 
 def predict(input_text):
     if classifier is None:
         return "Model not loaded properly.", {"Error": "Model not loaded properly."}
 
     try:
-        # Predict the category
+        input_text = normalize_input(input_text)
         result = classifier(input_text)
         label = result[0]["label"]
         score = result[0]["score"]
 
-        # Retrieve output and metadata dynamically from the dataset
         data = dataset_dict.get(input_text, {"output": "Unknown", "metadata": {}})
         output = data["output"]
         metadata = data["metadata"]
 
-        # Extract metadata details
         difficulty = metadata.get("difficulty", "Unknown")
         steps = metadata.get("steps", ["No steps available"])
 
-        # Create a simple result string without dashes
-        steps_text = "\n".join(steps)  # No dash or prefix for each step
+        steps_text = "\n".join(steps)
         simple_result = (
             f"Category: {label}\n"
             f"Confidence: {score:.2f}\n"
@@ -51,7 +50,6 @@ def predict(input_text):
             f"Steps:\n{steps_text}"
         )
 
-        # Create the full JSON output
         detailed_result = {
             "Category": label,
             "Confidence": score,
@@ -66,7 +64,6 @@ def predict(input_text):
 # Gradio interface
 import gradio as gr
 
-# Define sample inputs
 sample_inputs = [
     ["1 + 1 = ?"],
     ["(5 + 3) × 2 = ?"],
@@ -78,13 +75,12 @@ interface = gr.Interface(
     fn=predict,
     inputs=gr.Textbox(lines=2, placeholder="Enter a math problem..."),
     outputs=[
-        gr.Textbox(label="Simple Output"),  # Display only the result
-        gr.JSON(label="Detailed JSON Output"),  # Display full JSON
+        gr.Textbox(label="Simple Output"),
+        gr.JSON(label="Detailed JSON Output"),
     ],
     title="Sangapac Math Model",
     description="A model to classify math problems into categories like Arithmetic, Multiplication, Division, Algebra, and Geometry.",
-    examples=sample_inputs,  # Add examples below the Clear and Submit buttons
+    examples=sample_inputs,
 )
 
-# Launch the app
 interface.launch()