Update app.py

app.py

@@ -17,7 +17,6 @@ import matplotlib.pyplot as plt
 import seaborn as sns
 import numpy as np
 
-# Hugging Face transformers pipeline import
 from transformers import pipeline
 
 weather_map = {"Cloudy": 0, "Rainy": 1, "Sunny": 2}
@@ -36,11 +35,11 @@ try:
         df = pd.DataFrame(default_data)
     else:
         df = pd.read_csv("new_delay_data.csv")
-    
+
     df = pd.get_dummies(df, columns=["Phase"], drop_first=True)
     df["Weather"] = df["Weather"].map(weather_map)
     df.dropna(subset=["Weather", "Absentee", "DelayLog", "Delay%"], inplace=True)
-    
+
     for col in ["Phase_Framing", "Phase_Foundation"]:
         if col not in df.columns:
             df[col] = 0
@@ -65,9 +64,9 @@ except Exception as e:
     print(f"Error training model: {e}")
     raise
 
-print("Loading AI text generation model (EleutherAI/gpt-neo-125M)...")
+print("Loading AI text generation model (distilgpt2)...")
 try:
-    text_generator = pipeline("text-generation", model="EleutherAI/gpt-neo-125M")
+    text_generator = pipeline("text-generation", model="distilgpt2")
     print("Text generation model loaded successfully.")
 except Exception as e:
     print(f"Failed to load text generation model: {e}")
@@ -84,7 +83,8 @@ def generate_ai_insight(phase, weather, absentee_pct, delay_log, prediction):
         f"Provide a brief insight on the delay risks and a practical migration plan to mitigate these delays."
     )
     try:
-        result = text_generator(prompt, max_length=120, num_return_sequences=1, temperature=0.7, top_p=0.9)
+        result = text_generator(prompt, max_length=80, num_return_sequences=1,
+                                temperature=0.7, top_p=0.9, early_stopping=True)
         generated_text = result[0]['generated_text']
         insight = generated_text[len(prompt):].strip()
         return insight
@@ -99,20 +99,20 @@ def generate_heatmap(phase, weather, model):
         framing = 1 if phase == "Framing" else 0
         foundation = 1 if phase == "Foundation" else 0
         weather_encoded = weather_map.get(weather, 0)
-        
+
         Z = np.zeros((len(delay_log_range), len(absentee_range)))
         for i, delay_log in enumerate(delay_log_range):
             for j, absentee in enumerate(absentee_range):
                 input_data = [[framing, foundation, weather_encoded, absentee, delay_log]]
                 Z[i, j] = model.predict(input_data)[0]
-        
+
         plt.figure(figsize=(8, 6))
         sns.heatmap(Z, xticklabels=np.round(absentee_range, 1), yticklabels=np.round(delay_log_range, 1),
                     cmap="YlOrRd", annot=True, fmt=".1f", cbar_kws={'label': 'Predicted Delay %'})
         plt.xlabel("Absentee %")
         plt.ylabel("Previous Delay Log")
         plt.title(f"Delay Prediction Heatmap (Phase: {phase}, Weather: {weather})")
-        
+
         output_dir = "pdf_reports"
         os.makedirs(output_dir, exist_ok=True)
         timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
@@ -128,7 +128,7 @@ def generate_pdf_report(phase, weather, absentee_pct, delay_log, prediction, ris
     try:
         buffer = BytesIO()
         c = canvas.Canvas(buffer, pagesize=letter)
-        
+
         try:
             pdfmetrics.registerFont(TTFont('DejaVuSans', 'DejaVuSans.ttf'))
             c.setFont("DejaVuSans", 12)
@@ -137,13 +137,13 @@ def generate_pdf_report(phase, weather, absentee_pct, delay_log, prediction, ris
             print(f"Failed to load DejaVuSans font: {e}. Falling back to Helvetica with text flag.")
             c.setFont("Helvetica", 12)
             flag_indicator = " [FLAG]" if prediction >= 75 else ""
-        
+
         c.drawString(100, 750, "Project Delay Prediction Report")
         c.drawString(100, 730, f"Generated on: {datetime.now().strftime('%Y-%m-%d %H:%M:%S')}")
-        
+
         y_position = 700
         max_width = 400
-        
+
         details = [
             f"Phase: {phase}",
             f"Weather: {weather}",
@@ -153,38 +153,38 @@ def generate_pdf_report(phase, weather, absentee_pct, delay_log, prediction, ris
             f"Risk Level: {risk}",
             "AI Insight & Migration Plan:"
         ]
-        
+
         for line in details:
             lines = simpleSplit(line, 'DejaVuSans' if 'DejaVuSans' in pdfmetrics.getRegisteredFontNames() else 'Helvetica', 12, max_width)
             for wrapped_line in lines:
                 c.drawString(100, y_position, wrapped_line)
                 y_position -= 16
-        
+
         insight_lines = simpleSplit(insight, 'DejaVuSans' if 'DejaVuSans' in pdfmetrics.getRegisteredFontNames() else 'Helvetica', 12, max_width)
         for wrapped_line in insight_lines:
             c.drawString(100, y_position, wrapped_line)
             y_position -= 16
-        
+
         heatmap_path = generate_heatmap(phase, weather, model)
         if heatmap_path and os.path.exists(heatmap_path):
             c.drawString(100, y_position - 20, "Delay Prediction Heatmap:")
             c.drawImage(heatmap_path, 100, y_position - 250, width=400, height=200)
             y_position -= 270
-        
+
         c.showPage()
         c.save()
-        
+
         pdf_data = buffer.getvalue()
         buffer.close()
         pdf_base64 = base64.b64encode(pdf_data).decode("utf-8")
-        
+
         output_dir = "pdf_reports"
         os.makedirs(output_dir, exist_ok=True)
         timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
         output_path = os.path.join(output_dir, f"delay_report_{timestamp}.pdf")
         with open(output_path, "wb") as f:
             f.write(pdf_data)
-        
+
         return pdf_base64, output_path, heatmap_path
     except Exception as e:
         print(f"PDF generation failed: {e}")
@@ -274,7 +274,7 @@ try:
                     f"PDF Report: {'Saved locally at ' + pdf_path if pdf_path else 'Failed to generate'}\n"
                     f"Heatmap: {'Saved locally at ' + heatmap_path if heatmap_path else 'Failed to generate'}\n"
                     f"PDF Base64: {'Generated' if pdf_base64 else 'Not generated'}")
-        
+
         submit.click(
             predict_and_format,
             inputs=[phase_input, weather_input, absentee_input, delay_input],