Update utils.py

utils.py

@@ -9,7 +9,7 @@ import spaces
 import gc
 import time
 import random
-from chronos import ChronosPipeline # Menggunakan ChronosPipeline untuk Chronos-2
+from chronos import ChronosPipeline
 from scipy.stats import skew, kurtosis
 from typing import Dict, Union, List
 
@@ -44,15 +44,14 @@ def load_pipeline():
             clear_gpu_memory()
             print(f"Loading Chronos model: {model_name}...")
             
-            # PENTING: Optimasi untuk Chronos-2
+            # FIX 1: Menyederhanakan argumen untuk menghindari error 'input_patch_size'
             pipeline = ChronosPipeline.from_pretrained(
                 model_name,
                 device_map="cuda",  
                 torch_dtype=torch.float16, 
-                low_cpu_mem_usage=True,
-                trust_remote_code=True,
-                use_safetensors=True
+                # Menghapus argumen yang mungkin memicu error konfigurasi
             )
+            
             pipeline.model = pipeline.model.eval()
             for param in pipeline.model.parameters():
                 param.requires_grad = False
@@ -61,6 +60,7 @@ def load_pipeline():
         return pipeline
         
     except Exception as e:
+        # Menampilkan error yang lebih spesifik
         print(f"Error loading pipeline on CUDA, trying CPU: {str(e)}")
         try:
             # Fallback ke CPU
@@ -71,6 +71,8 @@ def load_pipeline():
         except Exception as cpu_e:
             raise RuntimeError(f"Failed to load model {model_name} on both CUDA and CPU: {str(cpu_e)}")
 
+# ... (Fungsi-fungsi lain: retry_yfinance_request, fetch_enhanced_covariates, calculate_advanced_risk_metrics)
+
 def retry_yfinance_request(func, max_retries=3, initial_delay=1):
     """Mekanisme retry untuk permintaan yfinance dengan backoff eksponensial."""
     for attempt in range(max_retries):
@@ -176,6 +178,10 @@ def predict_technical_indicators_future(data: pd.DataFrame, price_prediction: np
     """Memprediksi MACD dan Bollinger Bands di masa depan berdasarkan prediksi harga."""
     predictions = {}
     
+    # Pastikan price_prediction tidak kosong sebelum diolah
+    if price_prediction.size == 0:
+        return {"MACD_Future": np.array([]), "MACD_Signal_Future": np.array([]), "BB_Upper_Future": np.array([]), "BB_Lower_Future": np.array([])}
+
     full_price_series = np.concatenate([data['Close'].values, price_prediction])
     full_price_series = pd.Series(full_price_series)
     
@@ -208,8 +214,19 @@ def predict_technical_indicators_future(data: pd.DataFrame, price_prediction: np
 @spaces.GPU(duration=120)
 def predict_prices(data, prediction_days=30):
     """Fungsi prediksi utama menggunakan Chronos-2 dengan enhanced covariates."""
+    
+    # Default return structure for errors (FIX 2: Menggunakan np.array([]) yang aman)
+    empty_result = {
+        'values': np.array([]), 'dates': pd.Series([], dtype='datetime64[ns]'), 
+        'high_30d': 0, 'low_30d': 0, 'mean_30d': 0, 'change_pct': 0, 
+        'q01': np.array([]), 'q09': np.array([]),
+        'future_macd': np.array([]), 'future_macd_signal': np.array([]), 
+        'future_bb_upper': np.array([]), 'future_bb_lower': np.array([]),
+        'summary': 'Prediction failed due to model or data error.'
+    }
+    
     try:
-        # 1. Load Model
+        # 1. Load Model (Akan memanggil load_pipeline yang sudah diperbaiki)
         pipeline = load_pipeline()
         
         data_original = data.copy()
@@ -265,21 +282,23 @@ def predict_prices(data, prediction_days=30):
             'change_pct': change_pct, 
             'q01': q01_forecast,
             'q09': q09_forecast,
-            'future_macd': future_indicators.get('MACD_Future', []),
-            'future_macd_signal': future_indicators.get('MACD_Signal_Future', []),
-            'future_bb_upper': future_indicators.get('BB_Upper_Future', []),
-            'future_bb_lower': future_indicators.get('BB_Lower_Future', []),
+            'future_macd': future_indicators.get('MACD_Future', np.array([])),
+            'future_macd_signal': future_indicators.get('MACD_Signal_Future', np.array([])),
+            'future_bb_upper': future_indicators.get('BB_Upper_Future', np.array([])),
+            'future_bb_lower': future_indicators.get('BB_Lower_Future', np.array([])),
             'summary': f"AI Model: Amazon Chronos-2 (Enhanced Covariates: {len(all_covariates)} features)\nExpected High: {predicted_high:.2f}\nExpected Low: {predicted_low:.2f}\nExpected Change: {change_pct:.2f}%"
         }
         
     except Exception as e:
         error_message = f'Model prediction failed: {e}'
         print(f"Error in prediction: {e}")
-        return {'values': [], 'dates': [], 'high_30d': 0, 'low_30d': 0, 'mean_30d': 0, 'change_pct': 0, 'summary': error_message, 'q01': [], 'q09': [], 'future_macd': [], 'future_macd_signal': [], 'future_bb_upper': [], 'future_bb_lower': []}
+        empty_result['summary'] = error_message
+        return empty_result
 
 # Memperbarui fungsi create_prediction_chart untuk menampilkan Quantile Bands (q01, q09) dan Future BB
 def create_prediction_chart(data, predictions):
-    if not len(predictions['values']) or not len(predictions['q01']):
+    # Cek yang lebih aman untuk array kosong
+    if not predictions['values'].size or not predictions['q01'].size:
         return go.Figure().update_layout(title="Prediction Failed: No Data Available")
         
     fig = make_subplots(rows=2, cols=1, shared_xaxes=True, vertical_spacing=0.05, 
@@ -299,7 +318,7 @@ def create_prediction_chart(data, predictions):
     fig.add_trace(go.Scatter(x=predictions['dates'], y=predictions['values'], name='Median Forecast (Q0.5)', line=dict(color='red', width=3, dash='solid')), row=1, col=1)
 
     # Future Bollinger Bands 
-    if len(predictions['future_bb_upper']) == len(predictions['dates']):
+    if predictions['future_bb_upper'].size == predictions['dates'].size:
         fig.add_trace(go.Scatter(x=predictions['dates'], y=predictions['future_bb_upper'], name='BB Upper (Future)', line=dict(color='green', width=1, dash='dot')), row=1, col=1)
         fig.add_trace(go.Scatter(x=predictions['dates'], y=predictions['future_bb_lower'], name='BB Lower (Future)', line=dict(color='green', width=1, dash='dot')), row=1, col=1)
 
@@ -308,14 +327,16 @@ def create_prediction_chart(data, predictions):
     fig.add_trace(go.Scatter(x=[last_hist_date], y=[last_hist_price], mode='markers', marker=dict(size=10, color='blue', symbol='circle'), name='Last Known Price'), row=1, col=1)
 
     # 2. MACD Forecast (Row 2)
-    if len(predictions['future_macd']) == len(predictions['dates']):
+    if predictions['future_macd'].size == predictions['dates'].size:
         
+        # Perluas data historis MACD untuk charting yang lebih baik
+        lookback_period = 60
         macd_hist = data['Close'].ewm(span=12).mean() - data['Close'].ewm(span=26).mean()
         macd_signal_hist = macd_hist.ewm(span=9).mean()
         
-        macd_full = np.concatenate([macd_hist.iloc[-60:].values, predictions['future_macd']])
-        macd_signal_full = np.concatenate([macd_signal_hist.iloc[-60:].values, predictions['future_macd_signal']])
-        macd_dates_full = pd.to_datetime(np.concatenate([data.index[-60:].values, predictions['dates']]))
+        macd_full = np.concatenate([macd_hist.iloc[-lookback_period:].values, predictions['future_macd']])
+        macd_signal_full = np.concatenate([macd_signal_hist.iloc[-lookback_period:].values, predictions['future_macd_signal']])
+        macd_dates_full = pd.to_datetime(np.concatenate([data.index[-lookback_period:].values, predictions['dates']]))
         
         fig.add_trace(go.Scatter(x=macd_dates_full, y=macd_full, name='MACD Line', line=dict(color='blue', width=2)), row=2, col=1)
         fig.add_trace(go.Scatter(x=macd_dates_full, y=macd_signal_full, name='Signal Line', line=dict(color='red', width=1)), row=2, col=1)
@@ -334,9 +355,8 @@ def create_prediction_chart(data, predictions):
 
     return fig
 
-# --- Fungsi lama yang harus tetap ada ---
+# ... (Fungsi-fungsi lama lainnya seperti get_indonesian_stocks, calculate_technical_indicators, dll. tetap sama)
 def get_indonesian_stocks():
-    # ... (kode yang sama)
     return {
         "BBCA.JK": "Bank Central Asia", "BBRI.JK": "Bank BRI", "BBNI.JK": "Bank BNI",
         "BMRI.JK": "Bank Mandiri", "TLKM.JK": "Telkom Indonesia", "UNVR.JK": "Unilever Indonesia",
@@ -348,7 +368,6 @@ def get_indonesian_stocks():
     }
 
 def calculate_technical_indicators(data):
-    # Disesuaikan agar dapat menambahkan RSI, MACD, Signal ke DataFrame
     indicators = {}
     
     def calculate_rsi(prices, period=14):
@@ -391,7 +410,6 @@ def calculate_technical_indicators(data):
     indicators['volume'] = {'current': data['Volume'].iloc[-1], 'avg_20': data['Volume'].rolling(20).mean().iloc[-1], 'ratio': data['Volume'].iloc[-1] / data['Volume'].rolling(20).mean().iloc[-1]}
     
     # Tambahkan kolom indikator ke DataFrame input untuk digunakan nanti (di predict_technical_indicators_future)
-    # Catatan: Perubahan ini memodifikasi 'data' in-place.
     data['RSI'] = rsi_series
     data['MACD'] = macd
     data['MACD_Signal'] = signal_line
@@ -399,7 +417,6 @@ def calculate_technical_indicators(data):
     return indicators
 
 def generate_trading_signals(data, indicators):
-    # ... (kode yang sama)
     signals = {}
     current_price = data['Close'].iloc[-1]
     buy_signals = 0
@@ -458,7 +475,6 @@ def generate_trading_signals(data, indicators):
     return signals
 
 def get_fundamental_data(stock):
-    # ... (kode yang sama)
     try:
         info = stock.info
         history = stock.history(period="1d")
@@ -468,7 +484,6 @@ def get_fundamental_data(stock):
         return {'name': 'N/A', 'current_price': 0, 'market_cap': 0, 'pe_ratio': 0, 'dividend_yield': 0, 'volume': 0, 'info': 'Unable to fetch fundamental data'}
 
 def format_large_number(num):
-    # ... (kode yang sama)
     if num >= 1e12:
         return f"{num/1e12:.2f}T"
     elif num >= 1e9:
@@ -481,7 +496,6 @@ def format_large_number(num):
         return f"{num:.2f}"
 
 def create_price_chart(data, indicators):
-    # ... (kode yang sama)
     fig = make_subplots(rows=3, cols=1, shared_xaxes=True, vertical_spacing=0.05)
     fig.add_trace(go.Candlestick(x=data.index, open=data['Open'], high=data['High'], low=data['Low'], close=data['Close'], name='Price'), row=1, col=1)
     fig.add_trace(go.Scatter(x=data.index, y=indicators['moving_averages']['sma_20_values'], name='SMA 20', line=dict(color='orange')), row=1, col=1)
@@ -493,7 +507,6 @@ def create_price_chart(data, indicators):
     return fig
 
 def create_technical_chart(data, indicators):
-    # ... (kode yang sama)
     fig = make_subplots(rows=2, cols=2, subplot_titles=('Bollinger Bands', 'Volume', 'Price vs MA', 'RSI Analysis'))
     fig.add_trace(go.Scatter(x=data.index, y=data['Close'], name='Price', line=dict(color='black')), row=1, col=1)
     fig.add_trace(go.Scatter(x=data.index, y=indicators['bollinger']['upper_values'], name='Upper Band', line=dict(color='red')), row=1, col=1)