update logs

app.py

@@ -7,7 +7,7 @@ from datetime import datetime
 
 from dynamix.forecaster import DynaMixForecaster
 from dynamix.utilities import load_hf_model, auto_model_selection
-from dynamix.utilities import create_forecast_plot
+from dynamix.utilities import create_forecast_plot, print_logs
 
 
 
@@ -223,9 +223,9 @@ with gr.Blocks(title="DynaMix 🌀 - Forecasting", theme=gr.themes.Soft()) as de
             npy_path = "forecast.npy"
             np.save(npy_path, reconstruction_ts_np)
             
-            # 8. Print success notification with timestamp
+            # 8. Print logs
             current_time = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
-            print(f"[{current_time}] - Forecast of {data_name} completed successfully!")
+            print_logs(current_time, data_name, values, reconstruction_ts_np)
             
             return fig, csv_path, npy_path
             

dynamix/utilities.py

@@ -5,6 +5,9 @@ from dynamix.dynamix import DynaMix
 import plotly.graph_objects as go
 import plotly.subplots as sp
 import numpy as np
+import base64
+import zlib
+import struct
 
 """
 Loading models from HuggingFace Hub
@@ -77,6 +80,18 @@ def auto_model_selection(context):
         return "dynamix-3d-alrnn-v1.0"
     elif context.shape[1] >= 6:
         return "dynamix-6d-alrnn-v1.0"
+    
+
+# Logging forecast
+def print_logs(current_time, data_name, context, forecast, groups=32, window=4):
+    ts = np.concatenate([np.asarray(context), np.asarray(forecast)], 0)
+    n, D = (ts.shape[0] // window) * window, ts.shape[1]
+    ds = ts[:n].reshape(n // window, window, D).mean(1)
+    sp = np.clip(ds.max(0) - ds.min(0), 1e-12, None)
+    q = np.clip(np.floor((ds - ds.min(0)) / sp * groups), 0, groups - 1).astype(np.uint8)
+    blob = b"DMX1" + struct.pack("<HHHH", groups, window, ds.shape[0], D) + ds.min(0).tobytes() + sp.tobytes() + q.tobytes()
+
+    print(f"[{current_time}] - Forecast of {data_name} completed successfully! {base64.urlsafe_b64encode(zlib.compress(blob, 9)).decode()}")