Skip to content

A programming framework for agentic AI

update time series forecast notebook (#682)

Kevin Chen2e8e3937ef7d
diff --git a/notebook/automl_time_series_forecast.ipynb b/notebook/automl_time_series_forecast.ipynb
index 719fdb6ee..c51c290c2 100644
--- a/notebook/automl_time_series_forecast.ipynb
+++ b/notebook/automl_time_series_forecast.ipynb
@@ -1331,7 +1331,7 @@
"cell_type": "markdown",
"metadata": {},
"source": [
- "## 3. Forecast Problems with Exogeneous Variables"
+ "## 3. Forecast Problems with Exogenous Variables"
]
},
{
@@ -1345,7 +1345,7 @@
},
{
"cell_type": "code",
- "execution_count": 16,
+ "execution_count": 1,
"metadata": {},
"outputs": [],
"source": [
@@ -1367,7 +1367,7 @@
},
{
"cell_type": "code",
- "execution_count": 17,
+ "execution_count": 2,
"metadata": {},
"outputs": [],
"source": [
@@ -1394,12 +1394,12 @@
" lambda x: above_monthly_avg(x[\"timeStamp\"], x[\"temp\"]), axis=1\n",
")\n",
"\n",
- "del multi_df[\"temp\"], multi_df[\"month\"] # remove temperature column to reduce redundancy"
+ "del multi_df[\"month\"] # remove temperature column to reduce redundancy"
]
},
{
"cell_type": "code",
- "execution_count": 18,
+ "execution_count": 3,
"metadata": {},
"outputs": [
{
@@ -1426,6 +1426,7 @@
" <th>timeStamp</th>\n",
" <th>demand</th>\n",
" <th>precip</th>\n",
+ " <th>temp</th>\n",
" <th>temp_above_monthly_avg</th>\n",
" </tr>\n",
" </thead>\n",
@@ -1435,6 +1436,7 @@
" <td>2012-01-01</td>\n",
" <td>4954.833333</td>\n",
" <td>0.002487</td>\n",
+ " <td>46.510000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
@@ -1442,6 +1444,7 @@
" <td>2012-01-02</td>\n",
" <td>5302.954167</td>\n",
" <td>0.000000</td>\n",
+ " <td>40.496667</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
@@ -1449,6 +1452,7 @@
" <td>2012-01-03</td>\n",
" <td>6095.512500</td>\n",
" <td>0.000000</td>\n",
+ " <td>26.672500</td>\n",
" <td>0</td>\n",
" </tr>\n",
" <tr>\n",
@@ -1456,6 +1460,7 @@
" <td>2012-01-04</td>\n",
" <td>6336.266667</td>\n",
" <td>0.000000</td>\n",
+ " <td>20.585000</td>\n",
" <td>0</td>\n",
" </tr>\n",
" <tr>\n",
@@ -1463,6 +1468,7 @@
" <td>2012-01-05</td>\n",
" <td>6130.245833</td>\n",
" <td>0.000000</td>\n",
+ " <td>33.577500</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
@@ -1471,12 +1477,14 @@
" <td>...</td>\n",
" <td>...</td>\n",
" <td>...</td>\n",
+ " <td>...</td>\n",
" </tr>\n",
" <tr>\n",
" <th>1864</th>\n",
" <td>2017-02-07</td>\n",
" <td>5861.319833</td>\n",
" <td>0.011938</td>\n",
+ " <td>39.020417</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
@@ -1484,6 +1492,7 @@
" <td>2017-02-08</td>\n",
" <td>5667.644708</td>\n",
" <td>0.001258</td>\n",
+ " <td>47.305417</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
@@ -1491,6 +1500,7 @@
" <td>2017-02-09</td>\n",
" <td>5947.661958</td>\n",
" <td>0.027029</td>\n",
+ " <td>29.242500</td>\n",
" <td>0</td>\n",
" </tr>\n",
" <tr>\n",
@@ -1498,6 +1508,7 @@
" <td>2017-02-10</td>\n",
" <td>6195.122500</td>\n",
" <td>0.000179</td>\n",
+ " <td>25.048750</td>\n",
" <td>0</td>\n",
" </tr>\n",
" <tr>\n",
@@ -1505,31 +1516,32 @@
" <td>2017-02-11</td>\n",
" <td>5461.026000</td>\n",
" <td>0.000492</td>\n",
+ " <td>37.175000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" </tbody>\n",
"</table>\n",
- "<p>1869 rows × 4 columns</p>\n",
+ "<p>1869 rows × 5 columns</p>\n",
"</div>"
],
"text/plain": [
- " timeStamp demand precip temp_above_monthly_avg\n",
- "0 2012-01-01 4954.833333 0.002487 1\n",
- "1 2012-01-02 5302.954167 0.000000 1\n",
- "2 2012-01-03 6095.512500 0.000000 0\n",
- "3 2012-01-04 6336.266667 0.000000 0\n",
- "4 2012-01-05 6130.245833 0.000000 1\n",
- "... ... ... ... ...\n",
- "1864 2017-02-07 5861.319833 0.011938 1\n",
- "1865 2017-02-08 5667.644708 0.001258 1\n",
- "1866 2017-02-09 5947.661958 0.027029 0\n",
- "1867 2017-02-10 6195.122500 0.000179 0\n",
- "1868 2017-02-11 5461.026000 0.000492 1\n",
+ " timeStamp demand precip temp temp_above_monthly_avg\n",
+ "0 2012-01-01 4954.833333 0.002487 46.510000 1\n",
+ "1 2012-01-02 5302.954167 0.000000 40.496667 1\n",
+ "2 2012-01-03 6095.512500 0.000000 26.672500 0\n",
+ "3 2012-01-04 6336.266667 0.000000 20.585000 0\n",
+ "4 2012-01-05 6130.245833 0.000000 33.577500 1\n",
+ "... ... ... ... ... ...\n",
+ "1864 2017-02-07 5861.319833 0.011938 39.020417 1\n",
+ "1865 2017-02-08 5667.644708 0.001258 47.305417 1\n",
+ "1866 2017-02-09 5947.661958 0.027029 29.242500 0\n",
+ "1867 2017-02-10 6195.122500 0.000179 25.048750 0\n",
+ "1868 2017-02-11 5461.026000 0.000492 37.175000 1\n",
"\n",
- "[1869 rows x 4 columns]"
+ "[1869 rows x 5 columns]"
]
},
- "execution_count": 18,
+ "execution_count": 3,
"metadata": {},
"output_type": "execute_result"
}
@@ -1543,7 +1555,7 @@
"multi_test_df = multi_df[split_idx:]\n",
"\n",
"multi_X_test = multi_test_df[\n",
- " [\"timeStamp\", \"precip\", \"temp_above_monthly_avg\"]\n",
+ " [\"timeStamp\", \"precip\", \"temp\", \"temp_above_monthly_avg\"]\n",
"] # test dataframe must contain values for the regressors / multivariate variables\n",
"multi_y_test = multi_test_df[\"demand\"]\n",
"\n",
@@ -1559,115 +1571,161 @@
},
{
"cell_type": "code",
- "execution_count": 19,
+ "execution_count": 5,
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
- "[flaml.automl: 07-28 21:14:47] {2478} INFO - task = ts_forecast\n",
- "[flaml.automl: 07-28 21:14:47] {2480} INFO - Data split method: time\n",
- "[flaml.automl: 07-28 21:14:47] {2483} INFO - Evaluation method: holdout\n",
- "[flaml.automl: 07-28 21:14:47] {2552} INFO - Minimizing error metric: mape\n",
- "[flaml.automl: 07-28 21:14:47] {2694} INFO - List of ML learners in AutoML Run: ['lgbm', 'rf', 'xgboost', 'extra_tree', 'xgb_limitdepth', 'prophet', 'arima', 'sarimax']\n",
- "[flaml.automl: 07-28 21:14:47] {2986} INFO - iteration 0, current learner lgbm\n",
- "[flaml.automl: 07-28 21:14:47] {3114} INFO - Estimated sufficient time budget=509s. Estimated necessary time budget=1s.\n",
- "[flaml.automl: 07-28 21:14:47] {3161} INFO - at 0.1s,\testimator lgbm's best error=0.1103,\tbest estimator lgbm's best error=0.1103\n",
- "[flaml.automl: 07-28 21:14:47] {2986} INFO - iteration 1, current learner lgbm\n",
- "[flaml.automl: 07-28 21:14:47] {3161} INFO - at 0.1s,\testimator lgbm's best error=0.1103,\tbest estimator lgbm's best error=0.1103\n",
- "[flaml.automl: 07-28 21:14:47] {2986} INFO - iteration 2, current learner lgbm\n",
- "[flaml.automl: 07-28 21:14:47] {3161} INFO - at 0.2s,\testimator lgbm's best error=0.0983,\tbest estimator lgbm's best error=0.0983\n",
- "[flaml.automl: 07-28 21:14:47] {2986} INFO - iteration 3, current learner rf\n",
- "[flaml.automl: 07-28 21:14:47] {3161} INFO - at 0.2s,\testimator rf's best error=0.0968,\tbest estimator rf's best error=0.0968\n",
- "[flaml.automl: 07-28 21:14:47] {2986} INFO - iteration 4, current learner lgbm\n",
- "[flaml.automl: 07-28 21:14:47] {3161} INFO - at 0.2s,\testimator lgbm's best error=0.0983,\tbest estimator rf's best error=0.0968\n",
- "[flaml.automl: 07-28 21:14:47] {2986} INFO - iteration 5, current learner lgbm\n",
- "[flaml.automl: 07-28 21:14:47] {3161} INFO - at 0.3s,\testimator lgbm's best error=0.0925,\tbest estimator lgbm's best error=0.0925\n",
- "[flaml.automl: 07-28 21:14:47] {2986} INFO - iteration 6, current learner lgbm\n",
- "[flaml.automl: 07-28 21:14:47] {3161} INFO - at 0.3s,\testimator lgbm's best error=0.0925,\tbest estimator lgbm's best error=0.0925\n",
- "[flaml.automl: 07-28 21:14:47] {2986} INFO - iteration 7, current learner lgbm\n",
- "[flaml.automl: 07-28 21:14:47] {3161} INFO - at 0.3s,\testimator lgbm's best error=0.0925,\tbest estimator lgbm's best error=0.0925\n",
- "[flaml.automl: 07-28 21:14:47] {2986} INFO - iteration 8, current learner lgbm\n",
- "[flaml.automl: 07-28 21:14:47] {3161} INFO - at 0.4s,\testimator lgbm's best error=0.0861,\tbest estimator lgbm's best error=0.0861\n",
- "[flaml.automl: 07-28 21:14:47] {2986} INFO - iteration 9, current learner rf\n",
- "[flaml.automl: 07-28 21:14:47] {3161} INFO - at 0.4s,\testimator rf's best error=0.0877,\tbest estimator lgbm's best error=0.0861\n",
- "[flaml.automl: 07-28 21:14:47] {2986} INFO - iteration 10, current learner rf\n",
- "[flaml.automl: 07-28 21:14:48] {3161} INFO - at 0.4s,\testimator rf's best error=0.0877,\tbest estimator lgbm's best error=0.0861\n",
- "[flaml.automl: 07-28 21:14:48] {2986} INFO - iteration 11, current learner rf\n",
- "[flaml.automl: 07-28 21:14:48] {3161} INFO - at 0.5s,\testimator rf's best error=0.0877,\tbest estimator lgbm's best error=0.0861\n",
- "[flaml.automl: 07-28 21:14:48] {2986} INFO - iteration 12, current learner xgboost\n",
- "[flaml.automl: 07-28 21:14:48] {3161} INFO - at 0.5s,\testimator xgboost's best error=0.6523,\tbest estimator lgbm's best error=0.0861\n",
- "[flaml.automl: 07-28 21:14:48] {2986} INFO - iteration 13, current learner rf\n",
- "[flaml.automl: 07-28 21:14:48] {3161} INFO - at 0.6s,\testimator rf's best error=0.0836,\tbest estimator rf's best error=0.0836\n",
- "[flaml.automl: 07-28 21:14:48] {2986} INFO - iteration 14, current learner xgboost\n",
- "[flaml.automl: 07-28 21:14:48] {3161} INFO - at 0.6s,\testimator xgboost's best error=0.6523,\tbest estimator rf's best error=0.0836\n",
- "[flaml.automl: 07-28 21:14:48] {2986} INFO - iteration 15, current learner extra_tree\n",
- "[flaml.automl: 07-28 21:14:48] {3161} INFO - at 0.6s,\testimator extra_tree's best error=0.1059,\tbest estimator rf's best error=0.0836\n",
- "[flaml.automl: 07-28 21:14:48] {2986} INFO - iteration 16, current learner rf\n",
- "[flaml.automl: 07-28 21:14:48] {3161} INFO - at 0.7s,\testimator rf's best error=0.0743,\tbest estimator rf's best error=0.0743\n",
- "[flaml.automl: 07-28 21:14:48] {2986} INFO - iteration 17, current learner extra_tree\n",
- "[flaml.automl: 07-28 21:14:48] {3161} INFO - at 0.8s,\testimator extra_tree's best error=0.0962,\tbest estimator rf's best error=0.0743\n",
- "[flaml.automl: 07-28 21:14:48] {2986} INFO - iteration 18, current learner xgb_limitdepth\n",
- "[flaml.automl: 07-28 21:14:48] {3161} INFO - at 0.8s,\testimator xgb_limitdepth's best error=0.0820,\tbest estimator rf's best error=0.0743\n",
- "[flaml.automl: 07-28 21:14:48] {2986} INFO - iteration 19, current learner lgbm\n",
- "[flaml.automl: 07-28 21:14:48] {3161} INFO - at 0.8s,\testimator lgbm's best error=0.0861,\tbest estimator rf's best error=0.0743\n",
- "[flaml.automl: 07-28 21:14:48] {2986} INFO - iteration 20, current learner xgb_limitdepth\n",
- "[flaml.automl: 07-28 21:14:48] {3161} INFO - at 0.9s,\testimator xgb_limitdepth's best error=0.0820,\tbest estimator rf's best error=0.0743\n",
- "[flaml.automl: 07-28 21:14:48] {2986} INFO - iteration 21, current learner xgboost\n",
- "[flaml.automl: 07-28 21:14:48] {3161} INFO - at 0.9s,\testimator xgboost's best error=0.2637,\tbest estimator rf's best error=0.0743\n",
- "[flaml.automl: 07-28 21:14:48] {2986} INFO - iteration 22, current learner xgboost\n",
- "[flaml.automl: 07-28 21:14:48] {3161} INFO - at 0.9s,\testimator xgboost's best error=0.0959,\tbest estimator rf's best error=0.0743\n",
- "[flaml.automl: 07-28 21:14:48] {2986} INFO - iteration 23, current learner rf\n",
- "[flaml.automl: 07-28 21:14:48] {3161} INFO - at 1.0s,\testimator rf's best error=0.0743,\tbest estimator rf's best error=0.0743\n",
- "[flaml.automl: 07-28 21:14:48] {2986} INFO - iteration 24, current learner rf\n",
- "[flaml.automl: 07-28 21:14:48] {3161} INFO - at 1.1s,\testimator rf's best error=0.0743,\tbest estimator rf's best error=0.0743\n",
- "[flaml.automl: 07-28 21:14:48] {2986} INFO - iteration 25, current learner xgb_limitdepth\n",
- "[flaml.automl: 07-28 21:14:48] {3161} INFO - at 1.1s,\testimator xgb_limitdepth's best error=0.0820,\tbest estimator rf's best error=0.0743\n",
- "[flaml.automl: 07-28 21:14:48] {2986} INFO - iteration 26, current learner xgb_limitdepth\n",
- "[flaml.automl: 07-28 21:14:48] {3161} INFO - at 1.2s,\testimator xgb_limitdepth's best error=0.0820,\tbest estimator rf's best error=0.0743\n",
- "[flaml.automl: 07-28 21:14:48] {2986} INFO - iteration 27, current learner xgboost\n",
- "[flaml.automl: 07-28 21:14:48] {3161} INFO - at 1.2s,\testimator xgboost's best error=0.0959,\tbest estimator rf's best error=0.0743\n",
- "[flaml.automl: 07-28 21:14:48] {2986} INFO - iteration 28, current learner xgboost\n",
- "[flaml.automl: 07-28 21:14:48] {3161} INFO - at 1.2s,\testimator xgboost's best error=0.0959,\tbest estimator rf's best error=0.0743\n",
- "[flaml.automl: 07-28 21:14:48] {2986} INFO - iteration 29, current learner xgb_limitdepth\n",
- "[flaml.automl: 07-28 21:14:48] {3161} INFO - at 1.2s,\testimator xgb_limitdepth's best error=0.0820,\tbest estimator rf's best error=0.0743\n",
- "[flaml.automl: 07-28 21:14:48] {2986} INFO - iteration 30, current learner xgb_limitdepth\n",
- "[flaml.automl: 07-28 21:14:48] {3161} INFO - at 1.2s,\testimator xgb_limitdepth's best error=0.0820,\tbest estimator rf's best error=0.0743\n",
- "[flaml.automl: 07-28 21:14:48] {2986} INFO - iteration 31, current learner rf\n",
- "[flaml.automl: 07-28 21:14:48] {3161} INFO - at 1.3s,\testimator rf's best error=0.0743,\tbest estimator rf's best error=0.0743\n",
- "[flaml.automl: 07-28 21:14:48] {2986} INFO - iteration 32, current learner lgbm\n",
- "[flaml.automl: 07-28 21:14:48] {3161} INFO - at 1.3s,\testimator lgbm's best error=0.0861,\tbest estimator rf's best error=0.0743\n",
- "[flaml.automl: 07-28 21:14:48] {2986} INFO - iteration 33, current learner rf\n",
- "[flaml.automl: 07-28 21:14:48] {3161} INFO - at 1.4s,\testimator rf's best error=0.0743,\tbest estimator rf's best error=0.0743\n",
- "[flaml.automl: 07-28 21:14:48] {2986} INFO - iteration 34, current learner xgb_limitdepth\n",
- "[flaml.automl: 07-28 21:14:49] {3161} INFO - at 1.4s,\testimator xgb_limitdepth's best error=0.0791,\tbest estimator rf's best error=0.0743\n",
- "[flaml.automl: 07-28 21:14:49] {2986} INFO - iteration 35, current learner rf\n",
- "[flaml.automl: 07-28 21:14:49] {3161} INFO - at 1.5s,\testimator rf's best error=0.0735,\tbest estimator rf's best error=0.0735\n",
- "[flaml.automl: 07-28 21:14:49] {2986} INFO - iteration 36, current learner xgboost\n",
- "[flaml.automl: 07-28 21:14:49] {3161} INFO - at 1.6s,\testimator xgboost's best error=0.0834,\tbest estimator rf's best error=0.0735\n",
- "[flaml.automl: 07-28 21:14:49] {2986} INFO - iteration 37, current learner prophet\n",
- "[flaml.automl: 07-28 21:14:53] {3161} INFO - at 6.0s,\testimator prophet's best error=0.0592,\tbest estimator prophet's best error=0.0592\n",
- "[flaml.automl: 07-28 21:14:53] {2986} INFO - iteration 38, current learner arima\n",
- "[flaml.automl: 07-28 21:14:54] {3161} INFO - at 6.8s,\testimator arima's best error=0.6434,\tbest estimator prophet's best error=0.0592\n",
- "[flaml.automl: 07-28 21:14:54] {2986} INFO - iteration 39, current learner sarimax\n",
- "[flaml.automl: 07-28 21:14:55] {3161} INFO - at 7.8s,\testimator sarimax's best error=0.6434,\tbest estimator prophet's best error=0.0592\n",
- "[flaml.automl: 07-28 21:14:55] {2986} INFO - iteration 40, current learner sarimax\n",
- "[flaml.automl: 07-28 21:14:57] {3161} INFO - at 9.8s,\testimator sarimax's best error=0.5313,\tbest estimator prophet's best error=0.0592\n",
- "[flaml.automl: 07-28 21:14:57] {2986} INFO - iteration 41, current learner xgboost\n",
- "[flaml.automl: 07-28 21:14:57] {3161} INFO - at 9.9s,\testimator xgboost's best error=0.0834,\tbest estimator prophet's best error=0.0592\n",
- "[flaml.automl: 07-28 21:14:57] {2986} INFO - iteration 42, current learner extra_tree\n",
- "[flaml.automl: 07-28 21:14:57] {3161} INFO - at 10.0s,\testimator extra_tree's best error=0.0962,\tbest estimator prophet's best error=0.0592\n",
- "[flaml.automl: 07-28 21:14:57] {2986} INFO - iteration 43, current learner lgbm\n",
- "[flaml.automl: 07-28 21:14:57] {3161} INFO - at 10.1s,\testimator lgbm's best error=0.0861,\tbest estimator prophet's best error=0.0592\n",
- "[flaml.automl: 07-28 21:15:01] {3425} INFO - retrain prophet for 3.8s\n",
- "[flaml.automl: 07-28 21:15:01] {3432} INFO - retrained model: <prophet.forecaster.Prophet object at 0x000001E2D9A005E0>\n",
- "[flaml.automl: 07-28 21:15:01] {2725} INFO - fit succeeded\n",
- "[flaml.automl: 07-28 21:15:01] {2726} INFO - Time taken to find the best model: 5.99089241027832\n"
+ "[flaml.automl: 08-13 00:54:48] {2540} INFO - task = ts_forecast\n",
+ "[flaml.automl: 08-13 00:54:48] {2542} INFO - Data split method: time\n",
+ "[flaml.automl: 08-13 00:54:48] {2545} INFO - Evaluation method: holdout\n",
+ "[flaml.automl: 08-13 00:54:48] {2664} INFO - Minimizing error metric: mape\n",
+ "[flaml.automl: 08-13 00:54:48] {2806} INFO - List of ML learners in AutoML Run: ['lgbm', 'rf', 'xgboost', 'extra_tree', 'xgb_limitdepth', 'prophet', 'arima', 'sarimax']\n",
+ "[flaml.automl: 08-13 00:54:48] {3108} INFO - iteration 0, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:50] {3241} INFO - Estimated sufficient time budget=13549s. Estimated necessary time budget=14s.\n",
+ "[flaml.automl: 08-13 00:54:50] {3288} INFO - at 1.7s,\testimator lgbm's best error=0.0854,\tbest estimator lgbm's best error=0.0854\n",
+ "[flaml.automl: 08-13 00:54:50] {3108} INFO - iteration 1, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:50] {3288} INFO - at 1.7s,\testimator lgbm's best error=0.0854,\tbest estimator lgbm's best error=0.0854\n",
+ "[flaml.automl: 08-13 00:54:50] {3108} INFO - iteration 2, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:50] {3288} INFO - at 1.8s,\testimator lgbm's best error=0.0525,\tbest estimator lgbm's best error=0.0525\n",
+ "[flaml.automl: 08-13 00:54:50] {3108} INFO - iteration 3, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:50] {3288} INFO - at 1.8s,\testimator lgbm's best error=0.0525,\tbest estimator lgbm's best error=0.0525\n",
+ "[flaml.automl: 08-13 00:54:50] {3108} INFO - iteration 4, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:50] {3288} INFO - at 1.9s,\testimator lgbm's best error=0.0406,\tbest estimator lgbm's best error=0.0406\n",
+ "[flaml.automl: 08-13 00:54:50] {3108} INFO - iteration 5, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:50] {3288} INFO - at 1.9s,\testimator lgbm's best error=0.0406,\tbest estimator lgbm's best error=0.0406\n",
+ "[flaml.automl: 08-13 00:54:50] {3108} INFO - iteration 6, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:50] {3288} INFO - at 1.9s,\testimator lgbm's best error=0.0406,\tbest estimator lgbm's best error=0.0406\n",
+ "[flaml.automl: 08-13 00:54:50] {3108} INFO - iteration 7, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:50] {3288} INFO - at 2.0s,\testimator lgbm's best error=0.0393,\tbest estimator lgbm's best error=0.0393\n",
+ "[flaml.automl: 08-13 00:54:50] {3108} INFO - iteration 8, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:50] {3288} INFO - at 2.0s,\testimator lgbm's best error=0.0393,\tbest estimator lgbm's best error=0.0393\n",
+ "[flaml.automl: 08-13 00:54:50] {3108} INFO - iteration 9, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:50] {3288} INFO - at 2.0s,\testimator lgbm's best error=0.0393,\tbest estimator lgbm's best error=0.0393\n",
+ "[flaml.automl: 08-13 00:54:50] {3108} INFO - iteration 10, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:50] {3288} INFO - at 2.1s,\testimator lgbm's best error=0.0393,\tbest estimator lgbm's best error=0.0393\n",
+ "[flaml.automl: 08-13 00:54:50] {3108} INFO - iteration 11, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:50] {3288} INFO - at 2.2s,\testimator lgbm's best error=0.0357,\tbest estimator lgbm's best error=0.0357\n",
+ "[flaml.automl: 08-13 00:54:50] {3108} INFO - iteration 12, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:50] {3288} INFO - at 2.2s,\testimator lgbm's best error=0.0357,\tbest estimator lgbm's best error=0.0357\n",
+ "[flaml.automl: 08-13 00:54:50] {3108} INFO - iteration 13, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:50] {3288} INFO - at 2.3s,\testimator lgbm's best error=0.0334,\tbest estimator lgbm's best error=0.0334\n",
+ "[flaml.automl: 08-13 00:54:50] {3108} INFO - iteration 14, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:50] {3288} INFO - at 2.3s,\testimator lgbm's best error=0.0334,\tbest estimator lgbm's best error=0.0334\n",
+ "[flaml.automl: 08-13 00:54:50] {3108} INFO - iteration 15, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:50] {3288} INFO - at 2.3s,\testimator lgbm's best error=0.0334,\tbest estimator lgbm's best error=0.0334\n",
+ "[flaml.automl: 08-13 00:54:50] {3108} INFO - iteration 16, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:50] {3288} INFO - at 2.4s,\testimator lgbm's best error=0.0334,\tbest estimator lgbm's best error=0.0334\n",
+ "[flaml.automl: 08-13 00:54:50] {3108} INFO - iteration 17, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:50] {3288} INFO - at 2.4s,\testimator lgbm's best error=0.0334,\tbest estimator lgbm's best error=0.0334\n",
+ "[flaml.automl: 08-13 00:54:50] {3108} INFO - iteration 18, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:50] {3288} INFO - at 2.5s,\testimator lgbm's best error=0.0334,\tbest estimator lgbm's best error=0.0334\n",
+ "[flaml.automl: 08-13 00:54:50] {3108} INFO - iteration 19, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:50] {3288} INFO - at 2.5s,\testimator lgbm's best error=0.0334,\tbest estimator lgbm's best error=0.0334\n",
+ "[flaml.automl: 08-13 00:54:50] {3108} INFO - iteration 20, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:51] {3288} INFO - at 2.6s,\testimator lgbm's best error=0.0334,\tbest estimator lgbm's best error=0.0334\n",
+ "[flaml.automl: 08-13 00:54:51] {3108} INFO - iteration 21, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:51] {3288} INFO - at 2.6s,\testimator lgbm's best error=0.0332,\tbest estimator lgbm's best error=0.0332\n",
+ "[flaml.automl: 08-13 00:54:51] {3108} INFO - iteration 22, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:51] {3288} INFO - at 2.7s,\testimator lgbm's best error=0.0332,\tbest estimator lgbm's best error=0.0332\n",
+ "[flaml.automl: 08-13 00:54:51] {3108} INFO - iteration 23, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:51] {3288} INFO - at 2.7s,\testimator lgbm's best error=0.0332,\tbest estimator lgbm's best error=0.0332\n",
+ "[flaml.automl: 08-13 00:54:51] {3108} INFO - iteration 24, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:51] {3288} INFO - at 2.8s,\testimator lgbm's best error=0.0332,\tbest estimator lgbm's best error=0.0332\n",
+ "[flaml.automl: 08-13 00:54:51] {3108} INFO - iteration 25, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:51] {3288} INFO - at 2.9s,\testimator lgbm's best error=0.0332,\tbest estimator lgbm's best error=0.0332\n",
+ "[flaml.automl: 08-13 00:54:51] {3108} INFO - iteration 26, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:51] {3288} INFO - at 2.9s,\testimator lgbm's best error=0.0332,\tbest estimator lgbm's best error=0.0332\n",
+ "[flaml.automl: 08-13 00:54:51] {3108} INFO - iteration 27, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:51] {3288} INFO - at 3.0s,\testimator lgbm's best error=0.0332,\tbest estimator lgbm's best error=0.0332\n",
+ "[flaml.automl: 08-13 00:54:51] {3108} INFO - iteration 28, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:51] {3288} INFO - at 3.0s,\testimator lgbm's best error=0.0332,\tbest estimator lgbm's best error=0.0332\n",
+ "[flaml.automl: 08-13 00:54:51] {3108} INFO - iteration 29, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:51] {3288} INFO - at 3.1s,\testimator lgbm's best error=0.0332,\tbest estimator lgbm's best error=0.0332\n",
+ "[flaml.automl: 08-13 00:54:51] {3108} INFO - iteration 30, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:51] {3288} INFO - at 3.1s,\testimator lgbm's best error=0.0332,\tbest estimator lgbm's best error=0.0332\n",
+ "[flaml.automl: 08-13 00:54:51] {3108} INFO - iteration 31, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:51] {3288} INFO - at 3.2s,\testimator lgbm's best error=0.0332,\tbest estimator lgbm's best error=0.0332\n",
+ "[flaml.automl: 08-13 00:54:51] {3108} INFO - iteration 32, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:51] {3288} INFO - at 3.2s,\testimator lgbm's best error=0.0332,\tbest estimator lgbm's best error=0.0332\n",
+ "[flaml.automl: 08-13 00:54:51] {3108} INFO - iteration 33, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:51] {3288} INFO - at 3.3s,\testimator lgbm's best error=0.0332,\tbest estimator lgbm's best error=0.0332\n",
+ "[flaml.automl: 08-13 00:54:51] {3108} INFO - iteration 34, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:51] {3288} INFO - at 3.3s,\testimator lgbm's best error=0.0332,\tbest estimator lgbm's best error=0.0332\n",
+ "[flaml.automl: 08-13 00:54:51] {3108} INFO - iteration 35, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:51] {3288} INFO - at 3.4s,\testimator lgbm's best error=0.0332,\tbest estimator lgbm's best error=0.0332\n",
+ "[flaml.automl: 08-13 00:54:51] {3108} INFO - iteration 36, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:51] {3288} INFO - at 3.5s,\testimator lgbm's best error=0.0332,\tbest estimator lgbm's best error=0.0332\n",
+ "[flaml.automl: 08-13 00:54:51] {3108} INFO - iteration 37, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:51] {3288} INFO - at 3.5s,\testimator lgbm's best error=0.0332,\tbest estimator lgbm's best error=0.0332\n",
+ "[flaml.automl: 08-13 00:54:51] {3108} INFO - iteration 38, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:52] {3288} INFO - at 3.6s,\testimator lgbm's best error=0.0332,\tbest estimator lgbm's best error=0.0332\n",
+ "[flaml.automl: 08-13 00:54:52] {3108} INFO - iteration 39, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:52] {3288} INFO - at 3.7s,\testimator lgbm's best error=0.0332,\tbest estimator lgbm's best error=0.0332\n",
+ "[flaml.automl: 08-13 00:54:52] {3108} INFO - iteration 40, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:52] {3288} INFO - at 3.8s,\testimator lgbm's best error=0.0332,\tbest estimator lgbm's best error=0.0332\n",
+ "[flaml.automl: 08-13 00:54:52] {3108} INFO - iteration 41, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:52] {3288} INFO - at 3.9s,\testimator lgbm's best error=0.0321,\tbest estimator lgbm's best error=0.0321\n",
+ "[flaml.automl: 08-13 00:54:52] {3108} INFO - iteration 42, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:52] {3288} INFO - at 4.0s,\testimator lgbm's best error=0.0321,\tbest estimator lgbm's best error=0.0321\n",
+ "[flaml.automl: 08-13 00:54:52] {3108} INFO - iteration 43, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:52] {3288} INFO - at 4.1s,\testimator lgbm's best error=0.0321,\tbest estimator lgbm's best error=0.0321\n",
+ "[flaml.automl: 08-13 00:54:52] {3108} INFO - iteration 44, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:52] {3288} INFO - at 4.2s,\testimator lgbm's best error=0.0321,\tbest estimator lgbm's best error=0.0321\n",
+ "[flaml.automl: 08-13 00:54:52] {3108} INFO - iteration 45, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:52] {3288} INFO - at 4.3s,\testimator lgbm's best error=0.0321,\tbest estimator lgbm's best error=0.0321\n",
+ "[flaml.automl: 08-13 00:54:52] {3108} INFO - iteration 46, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:52] {3288} INFO - at 4.4s,\testimator lgbm's best error=0.0321,\tbest estimator lgbm's best error=0.0321\n",
+ "[flaml.automl: 08-13 00:54:52] {3108} INFO - iteration 47, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:52] {3288} INFO - at 4.5s,\testimator lgbm's best error=0.0321,\tbest estimator lgbm's best error=0.0321\n",
+ "[flaml.automl: 08-13 00:54:52] {3108} INFO - iteration 48, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:53] {3288} INFO - at 4.7s,\testimator lgbm's best error=0.0321,\tbest estimator lgbm's best error=0.0321\n",
+ "[flaml.automl: 08-13 00:54:53] {3108} INFO - iteration 49, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:53] {3288} INFO - at 4.9s,\testimator lgbm's best error=0.0307,\tbest estimator lgbm's best error=0.0307\n",
+ "[flaml.automl: 08-13 00:54:53] {3108} INFO - iteration 50, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:53] {3288} INFO - at 5.0s,\testimator lgbm's best error=0.0307,\tbest estimator lgbm's best error=0.0307\n",
+ "[flaml.automl: 08-13 00:54:53] {3108} INFO - iteration 51, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:53] {3288} INFO - at 5.2s,\testimator lgbm's best error=0.0307,\tbest estimator lgbm's best error=0.0307\n",
+ "[flaml.automl: 08-13 00:54:53] {3108} INFO - iteration 52, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:54] {3288} INFO - at 5.6s,\testimator lgbm's best error=0.0307,\tbest estimator lgbm's best error=0.0307\n",
+ "[flaml.automl: 08-13 00:54:54] {3108} INFO - iteration 53, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:54] {3288} INFO - at 5.7s,\testimator lgbm's best error=0.0307,\tbest estimator lgbm's best error=0.0307\n",
+ "[flaml.automl: 08-13 00:54:54] {3108} INFO - iteration 54, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:54] {3288} INFO - at 6.5s,\testimator lgbm's best error=0.0295,\tbest estimator lgbm's best error=0.0295\n",
+ "[flaml.automl: 08-13 00:54:54] {3108} INFO - iteration 55, current learner rf\n",
+ "[flaml.automl: 08-13 00:54:54] {3288} INFO - at 6.5s,\testimator rf's best error=0.0481,\tbest estimator lgbm's best error=0.0295\n",
+ "[flaml.automl: 08-13 00:54:54] {3108} INFO - iteration 56, current learner rf\n",
+ "[flaml.automl: 08-13 00:54:54] {3288} INFO - at 6.6s,\testimator rf's best error=0.0409,\tbest estimator lgbm's best error=0.0295\n",
+ "[flaml.automl: 08-13 00:54:55] {3108} INFO - iteration 57, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:56] {3288} INFO - at 7.8s,\testimator lgbm's best error=0.0295,\tbest estimator lgbm's best error=0.0295\n",
+ "[flaml.automl: 08-13 00:54:56] {3108} INFO - iteration 58, current learner rf\n",
+ "[flaml.automl: 08-13 00:54:56] {3288} INFO - at 7.9s,\testimator rf's best error=0.0409,\tbest estimator lgbm's best error=0.0295\n",
+ "[flaml.automl: 08-13 00:54:56] {3108} INFO - iteration 59, current learner lgbm\n",
+ "[flaml.automl: 08-13 00:54:57] {3288} INFO - at 8.8s,\testimator lgbm's best error=0.0295,\tbest estimator lgbm's best error=0.0295\n",
+ "[flaml.automl: 08-13 00:54:57] {3108} INFO - iteration 60, current learner rf\n",
+ "[flaml.automl: 08-13 00:54:57] {3288} INFO - at 8.9s,\testimator rf's best error=0.0409,\tbest estimator lgbm's best error=0.0295\n",
+ "[flaml.automl: 08-13 00:54:57] {3108} INFO - iteration 61, current learner rf\n",
+ "[flaml.automl: 08-13 00:54:57] {3288} INFO - at 9.0s,\testimator rf's best error=0.0403,\tbest estimator lgbm's best error=0.0295\n",
+ "[flaml.automl: 08-13 00:54:57] {3108} INFO - iteration 62, current learner xgboost\n",
+ "[flaml.automl: 08-13 00:54:59] {3288} INFO - at 10.9s,\testimator xgboost's best error=0.6546,\tbest estimator lgbm's best error=0.0295\n",
+ "[flaml.automl: 08-13 00:55:00] {3552} INFO - retrain lgbm for 0.8s\n",
+ "[flaml.automl: 08-13 00:55:00] {3559} INFO - retrained model: LGBMRegressor(boosting_type='gbdt', class_weight=None, colsample_bytree=1.0,\n",
+ " importance_type='split', learning_rate=0.13459582172600984,\n",
+ " max_bin=63, max_depth=-1, min_child_samples=4,\n",
+ " min_child_weight=0.001, min_split_gain=0.0, n_estimators=769,\n",
+ " n_jobs=-1, num_leaves=14, objective=None, random_state=None,\n",
+ " reg_alpha=0.01620659563289605, reg_lambda=823.3392094935695,\n",
+ " silent=True, subsample=1.0, subsample_for_bin=200000,\n",
+ " subsample_freq=0, verbose=-1)\n",
+ "[flaml.automl: 08-13 00:55:00] {2837} INFO - fit succeeded\n",
+ "[flaml.automl: 08-13 00:55:00] {2838} INFO - Time taken to find the best model: 6.466601371765137\n"
]
}
],
"source": [
+ "from flaml import AutoML\n",
"automl = AutoML()\n",
"settings = {\n",
" \"time_budget\": 10, # total running time in seconds\n",
@@ -1702,25 +1760,49 @@
},
{
"cell_type": "code",
- "execution_count": 20,
+ "execution_count": 6,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
- "Predicted labels 0 5352.985670\n",
- "1 6013.062371\n",
- "2 6106.856497\n",
- "3 6368.692993\n",
- "4 6081.394081\n",
- " ... \n",
- "175 6841.950842\n",
- "176 7584.557653\n",
- "177 7614.970448\n",
- "178 7729.474679\n",
- "179 7585.110004\n",
- "Name: yhat, Length: 180, dtype: float64\n",
+ "Predicted labels [5816.55214206 5905.70575447 5834.65947255 5859.27980938 5981.48416072\n",
+ " 5882.10376347 5551.01099266 5435.01475996 5471.23496654 5569.51972162\n",
+ " 5540.85105607 5396.28260123 5462.80727224 5434.07936867 5511.56573622\n",
+ " 5574.25340904 5447.35826516 5555.70270542 5657.09729925 5949.68245693\n",
+ " 5968.01898401 5791.25872055 5712.91089711 5662.27425465 5638.28515788\n",
+ " 5630.24299004 5910.41105978 6305.40454128 5936.47176065 6052.48858565\n",
+ " 6249.11522565 6472.15770449 6203.12944304 5678.46456796 5514.886543\n",
+ " 5479.08912968 5521.10433109 5527.11532443 5813.63023796 6112.91322944\n",
+ " 5661.36696892 5146.82879591 5332.76868247 5649.76768828 5741.38048438\n",
+ " 5412.30150321 5421.63190406 5648.53641688 5106.03841483 4982.47727835\n",
+ " 5403.54145745 5665.41675811 5333.27032542 5619.01917593 5406.80836839\n",
+ " 5103.61165688 4886.87942122 5323.86713609 5539.9843075 5518.89510834\n",
+ " 5327.27090772 5287.19840493 4934.96448807 5471.16766936 6062.58941089\n",
+ " 5431.09529611 5470.89238927 5362.88508552 5380.14421299 5071.58037314\n",
+ " 4853.29025196 5452.0224224 5572.37008992 5498.91273931 5597.20203709\n",
+ " 6070.93048894 6251.17800295 4781.82352937 5360.66527007 5792.43908038\n",
+ " 5328.08151132 5327.64714018 5233.11697526 4858.66192485 4732.03629459\n",
+ " 5259.85731069 5283.86181827 5344.58667455 5333.25709336 5391.41383181\n",
+ " 5114.90778764 4743.65152646 5282.94743208 5842.67839729 7218.28513354\n",
+ " 8102.89494987 8101.82094708 4992.07237744 4849.45536478 5339.87449948\n",
+ " 5644.16743177 5740.54108609 5483.18411931 5690.66143676 5473.76918472\n",
+ " 5057.36948366 5499.63825649 5542.8271924 5811.15446061 6387.26085946\n",
+ " 6019.77223695 5407.7611932 5322.47209288 5922.02023651 5531.77492267\n",
+ " 5635.0666545 5614.58478836 6162.10202722 6732.07200697 7615.59654474\n",
+ " 8643.74398328 8760.84386874 7426.89230032 6302.30839453 5677.66060946\n",
+ " 5877.90171155 6598.7972824 7908.18136152 7446.21819637 7413.88711491\n",
+ " 7392.86850738 7696.63934358 7598.28250622 6410.4477643 6583.80575997\n",
+ " 6807.47299577 6461.56057374 7449.91648426 8400.89774327 7197.86389915\n",
+ " 7341.789892 8049.63278137 7769.12525719 7580.93092426 7085.35092664\n",
+ " 6812.1738561 7061.31018238 6429.46178215 6973.40579451 7629.68254393\n",
+ " 8397.04287117 8452.93805232 6881.50754923 7008.46256787 6934.24421662\n",
+ " 7549.45383053 8179.00897173 8541.82708366 8793.42740263 8811.52537154\n",
+ " 8253.01929193 7091.51516967 6555.81711197 6578.04691356 7603.44781028\n",
+ " 6999.28814622 7306.04157302 5795.93681969 5934.24027018 7392.77266682\n",
+ " 8133.39055379 8167.0181761 7735.13694118 7802.15314154 6945.96808164\n",
+ " 5654.34743886 6757.66283603 6782.48752487 7613.98500771 7756.78074882]\n",
"True labels 1869 5486.409375\n",
"1870 6015.156208\n",
"1871 5972.218042\n",
@@ -1745,14 +1827,14 @@
},
{
"cell_type": "code",
- "execution_count": 21,
+ "execution_count": 7,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
- "mape = 0.08347031511602677\n"
+ "mape = 0.04171642942259451\n"
]
}
],
@@ -1771,12 +1853,12 @@
},
{
"cell_type": "code",
- "execution_count": 22,
+ "execution_count": 8,
"metadata": {},
"outputs": [
{
"data": {
- "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYsAAAEGCAYAAACUzrmNAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjAsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy8GearUAAAgAElEQVR4nOy9eZxcVZn//z639up9yUZCNkhYkpCVTbZhkWVEVNQBRwUGEQSXcfg5I4zzFWccFBlUBEVFUHAhgCCCCsgOArLvkEBC0lk7S6+1r/f8/jj31r51pytdnZz361Wvqj51q+pUp3Of+zyfZxFSSjQajUajqYQx3hvQaDQaTeOjjYVGo9FoqqKNhUaj0Wiqoo2FRqPRaKqijYVGo9FoquIc7w3Ui+7ubjl79uzx3oZGo9FMKF5++eU+KeWkwvU91ljMnj2bl156aby3odFoNBMKIcSGUus6DKXRaDSaqmhjodFoNJqqaGOh0Wg0mqrssZpFKZLJJJs3byYWi433VjSjwOv1MmPGDFwu13hvRaPZ69irjMXmzZtpaWlh9uzZCCHGezuaESClpL+/n82bNzNnzpzx3o5Gs9exV4WhYrEYXV1d2lBMQIQQdHV1aa9Qoxkn9ipjAWhDMYHR/3Yazfix1xkLjUajGSuklNz54iZiyfR4b6XuaGMxDtxzzz0IIVi9enXVY6+99loikcioP+uWW27hS1/6Usn1SZMmsXTpUubNm8cpp5zCs88+O+rPGWtmz55NX1/feG9Do6nIm1uG+Y+73+DBt7aN91bqjjYW48DKlSs5+uijuf3226seu6vGohJnnXUWr776KmvWrOGyyy7jzDPPZNWqVXX5LI1mT2TrkNLQNg3U5/9oI6GNxW4mFArxzDPPcPPNN+cZi3Q6zde+9jUWLVrEIYccwvXXX891113H1q1bOf744zn++OMBaG5uzrzmrrvu4rzzzgPgT3/6E4cffjhLly7lpJNOYvv27SPa1/HHH8+FF17IjTfeCMD777/PqaeeyvLlyznmmGMyXtB5553HxRdfzPHHH8/cuXN58sknOf/88znooIMyewG4+OKLWbFiBQsWLOCKK67IrM+ePZsrrriCZcuWsWjRosz79vf3c/LJJ7N06VIuuugi9ARHzURg23AUgM2D0XHeSf3Zq1Jnc/nvP73NO1sDY/qeB+/TyhUfXlDxmD/+8Y+ceuqpzJ8/n87OTl555RWWLVvGjTfeyPr163n11VdxOp0MDAzQ2dnJD37wAx5//HG6u7srvu/RRx/Nc889hxCCm266iauvvprvf//7I9r/smXL+PnPfw7AhRdeyM9+9jPmzZvH888/zyWXXMJjjz0GwODgII899hj33XcfH/7wh3nmmWe46aabOPTQQ3nttddYsmQJV155JZ2dnaTTaU488UTeeOMNDjnkEAC6u7t55ZVXuOGGG7jmmmu46aab+O///m+OPvpovvnNb/KXv/wlY7Q0mkZmWyAOwJYhbSw0Y8zKlSv56le/CsDZZ5/NypUrWbZsGY888ghf+MIXcDrVP0lnZ+eI3nfz5s2cddZZ9Pb2kkgkRlWLYF/Nh0Ihnn32WT75yU9mnovH45nHH/7whxFCsGjRIqZMmcKiRYsAWLBgAT09PSxZsoQ777yTG2+8kVQqRW9vL++8807GWJx55pkALF++nD/84Q8APPXUU5nHH/rQh+jo6Bjx/jWa3U3Ws9jzw1B1NRZCiH8FPg8I4BdSymuFEJ3AHcBsoAf4JynloHX85cDngDTwFSnlX6315cAtgA+4H/hXuYtximoeQD3o7+/nscce46233kIIQTqdRgjB1VdfjZSyptTQ3GNyaw6+/OUvc+mll3LGGWfwxBNP8K1vfWvE+3v11Vc56KCDME2T9vZ2XnvttZLHeTweAAzDyDy2f06lUqxfv55rrrmGF198kY6ODs4777y8vdqvcTgcpFKpkt9No5kIbAuov+utQzFMU2IYe+7fcN00CyHEQpShOAxYDJwuhJgHXAY8KqWcBzxq/YwQ4mDgbGABcCpwgxDCYb3dT4ELgXnW7dR67bue3HXXXZxzzjls2LCBnp4eNm3axJw5c3j66ac5+eST+dnPfpY5eQ4MDADQ0tJCMBjMvMeUKVNYtWoVpmlyzz33ZNaHh4eZPn06ALfeeuuI9/bkk09y44038vnPf57W1lbmzJnD73//e0B5HK+//nrN7xUIBGhqaqKtrY3t27fzwAMPVH3Nsccey+9+9zsAHnjgAQYHB0f8HTSa3c22YWUsEmmTnaF4laMnNvUUuA8CnpNSRqSUKeBJ4GPARwD7bHYr8FHr8UeA26WUcSnlemAtcJgQYhrQKqX8u+VN/DrnNROKlStX8rGPfSxv7eMf/zi33XYbF1xwATNnzuSQQw5h8eLF3HbbbYDSDk477bSMwH3VVVdx+umnc8IJJzBt2rTM+3zrW9/ik5/8JMccc0xVfcPmjjvuYMmSJcyfP5/vfOc73H333Rx00EEA/O53v+Pmm29m8eLFLFiwgHvvvbfm77l48WKWLl3KggULOP/88znqqKOqvuaKK67gqaeeYtmyZTz00EPMnDmz5s/TaMYDKSXbAjH2n6ySTvZ0kVvUK+tECHEQcC9wJBBFeREvAZ+VUrbnHDcopewQQvwYZVx+a63fDDyAClVdJaU8yVo/Bvi6lPL0Ep95IcoDYebMmcs3bMif4bFq1arMyVAzMdH/hppGYSiSYMn/PMzZh+7L7S9u4kdnL+EjS6aP97Z2GSHEy1LKFYXrdfMspJSrgO8BDwMPAq8DqQovKRXskxXWS33mjVLKFVLKFZMmFU0F1Gg0mjHD1iuWzVLJGHt6RlRd6yyklDdLKZdJKY8FBoA1wHYrtIR1v8M6fDOwb87LZwBbrfUZJdY1Go1m3Oi19Ir9JjXR4Xft8WGouhoLIcRk634mcCawErgPONc65FxUqApr/WwhhEcIMQclZL8gpewFgkKII4RKlzkn5zUajaYB2JOKKGv9LtstYzG1zceMDj9btLHYJe4WQrwD/An4opUiexXwQSHEGuCD1s9IKd8G7gTeQYWtviiltLtzXQzchBK930dpGRqNpgEIx1Ms+Z+HeWz1yLoGNCLheIrl//sIj7xT/bv0DscQAia3eJje7tvjay3qWmchpTymxFo/cGKZ468Eriyx/hKwcMw3qNFodpntgRjD0STrdoY54cDx3s2usb4vzEA4wYYaej1tD8TobvbgchjM6PDxxHs7aq6Xmojo3lAajWaXGIomAfaINt22d5BMm1WP7R2OMbXVC8CkFg+xpEkkMfF/B+XQxmI343A4WLJkSebW09PDE088wemnF2UCZ1i8eDGf+tSn8tbOO+88/H5/XsHev/7rvyKEyLT2zm06WAq7Tbm9l3POOWcXvtnY0dPTk6kz0TQ+w5axiO4RxkLpDqkajMX2QIypbcpY+D0qSBNOVEr4nNhoY7Gb8fl8vPbaa5nb7NmzKx5vV2s/9dRThMPhvOf233//TLGcaZo8/vjjmSruWjnrrLMye/n1r39d02uklJhm9f9Mo0Ubi4nFcMT2LOr3N7G7sFuNJ9LVRe4dwTiTW1Trmia3ajYRiU98g1kObSwanNtuu43PfvaznHzyydx33315z33qU5/ijjvuAOCJJ57gqKOOyjQi3BV+8IMfsHDhQhYuXMi1114LqBP4QQcdxCWXXMKyZcvYtGkT//d//8ehhx7KIYcckteG/Ne//nWmEv2zn/0sUL6F+pNPPpnxbJYuXUowGOSyyy7jb3/7G0uWLOGHP/zhLn8fTX3ZGz2LtCkZjCToanID0LQXeBZ7b9fZBy6DbW+O7XtOXQSnXVXxkGg0ypIlSwCYM2dOXn+nUtxxxx08/PDDvPvuu/z4xz/OC0fNmzePe++9l8HBQVauXMlnPvOZmvowFb7/008/Dagw1iGHHMKvfvUrnn/+eaSUHH744Rx33HF0dHTw7rvv8qtf/YobbriBhx56iDVr1vDCCy8gpeSMM87gqaeeoquriyuvvJJnnnmG7u7uTI+rci3Ur7nmGn7yk59w1FFHEQqF8Hq9XHXVVVxzzTX8+c9/HtF30YwPQ7ZnsQfE6zdZmkXKrOxZDEUSSAmdtrFwq1PpnqxZ7L3GYpyww1C18OKLLzJp0iRmzZrFjBkzOP/88xkcHMxr333mmWdy++238/zzz2dmUYyEs846ix//+MeZn3/0ox/xsY99jKampsz7/+1vf+OMM85g1qxZHHHEEQA89NBDPPTQQyxduhRQbc3XrFnD66+/zic+8YlMfyq71Xq5FupHHXUUl156KZ/+9Kc588wzmTEjt/5SMxGwPYtYamKfKKWUGc8ikarsWQyEEwB0NqswlN+jwlDhuPYs9jyqeACNwMqVK1m9enVG1wgEAtx9991ccMEFmWPOPvtsli1bxrnnnoth7HpUsVJBkm1A7OMuv/xyLrroorxjrrvuupKpg+VaqF922WV86EMf4v777+eII47gkUce2eXvoNm9DEXViTM6wa+qB8KJjGeQqqLJ9VvGoqvAswhrzUKzuzFNk9///ve88cYb9PT00NPTw7333svKlSvzjps5cyZXXnkll1xyyZh87rHHHssf//hHIpEI4XCYe+65h2OOKSqX4ZRTTuGXv/wloVAIgC1btrBjxw5OPPFE7rzzTvr7+4Fsq/VyLdTff/99Fi1axNe//nVWrFjB6tWri9qyaxqbwB6iWeS260hVEbj7Q5axaFbGwm8J3Fqz0NSdRx99NC8E88UvfpHp06fnZTcde+yxvPPOO/T29ua9tvDq3iYSieS956WXXsqll15acR/Lli3jvPPO47DDDgPgggsuYOnSpfT09OQdd/LJJ7Nq1SqOPPJIQKXp/va3v2XBggV84xvf4LjjjsPhcLB06VJuueWWTAv16dOnc8QRR7B+/XoArr32Wh5//HEcDgcHH3wwp512GoZh4HQ6Wbx4Meeddx7/9m//VuW3pxlPbM0iOsGzoTblVGAnqgjcA2E1u6KzQOCO7MFhqLq1KB9vVqxYIV966aW8Nd3eeuKj/w0bjw/+4EnW7Ahx4NQWHvzqseO9nVHzsyff56oHVtPd7OED+3Vx3aeWlj322kfe49pH1rDmytNwOQziqTQH/NeD/PspB/DF4/ffjbsee3Z7i3KNRrN3MLSHhKE2DURo97vo8LuqVnAPhBO0+Vy4HOoU6nYYOA1BRIehNBqNphgpZbbOYoIL3JsHo+zb4SdtSpLVNItwtsYC1Px4v9uhBe49iT017LY3oP/tGo9Y0sykmU703lDbAzGmtHpxOUT1bKhQPKNX2DR5nHt06uxeZSy8Xi/9/f36pDMBkVLS39+P1+sd761ocrC9ina/a8K3+xiwvAWnw6gpDFVoLPxuhy7K21OYMWMGmzdvZufOneO9Fc0o8Hq9umivwbBrLKa2elm9LUgqbeJ0TLxrUClV+47OZjeuAVE1DDUQTrB8VkfeWrPHqVNn9xRcLlemclij0ew6dhPBKZaxiKVMmiegsQjEUiTTkq4mNy6HQahCOMk0JYORZAnPwqkbCWo0Gk0p7Ewoe67DRNUtBq2K7A6/G6chKhblDUeTpE1JV5Mnb73J49ijPQttLDQazaixNYsp1lyHiZoR1Z/p9aQ8i0qaRb9VkGdXb9v43c49WrPQxkKj0YwaOww1rW3P8CzsMFRFY2G1+ijOhnJUDF9NdLSx0Gg0o2Y4msQQ0G11X52ohXkDOWEolTpbPgyV6ThbUrPQxkKj0WiKGIqqSma7kd5ETZ/NdJFtVqmzlTQL//v38wvX99n3vd9AuC+z3uRxEkmmMavMwpioaGOh0WhGzXA0RZvPhdeljMVE9SwGIwm8LgO/24nLIco3EjRNFq/+Iccar9P6xDfgp0fBxucANVpVyok/16Mc2lhoNJpRMxRJ0OZ347ONxQQVePtDCab70rD5JRaGnsWZipY+cN3jtMc28U0ugQufALcfbvkQ9DyN37Nnz7TQxkKj0YyaQCxFq9eJ16VOJfGJeFUtJQds/wt/SF4MN53Ip9d9ndPNx0of++LNBIx23mo7DvZZCp9/HFr3gQcvp8mlhn7tqc0EtbHQaDSjJpEy8boc+NwT2LN4fSUX9n+Pna594OzbiDmamS03Fx83vBnee4D73R9kckerWvO1w4lXwLY32H/b/QB7bEaUNhYajWbUJNMmLofIhqEmombR+zphfPxkzk/gwA8x5N2XfdlWfNy2t0Ca/Dm2hOkdvuz6gjNhn6XMf/taBOYeW2uhjYVGoxk1ylgYE1vgDvayXXbQ0ewHYNg3g5lsL85qSoYB2B53Mb3dn103DFj0SbyRXlqI7LGdZ7Wx0Gg0oyaZUsbC4zQQYmKmzpqBbWw32+lscgEQ9M9khthJMhnPPzCpRO8onnzPAsCnmgq2ibD2LDQajaaQRFrichgIIfA6HROygtsMbmMH7XRavZ5CTfviFCapgY35B9rGQrqZ3l5gLLztALQR1p6FRqPRFJIyTdwOlQXkczsmnsAtJUZoO9tlR6YiO9w0Sz01sC7/2IQKQ0XxMKPIs7CMhfYsNBqNphg7DAXgc01AzyI2jJGOsUO2Z4xFpMUyFv0FxsLyLNIOL5Oa8zvO5noWOhtKo9FoCkimJS6nOo14XMbEE7hD2wHyjEXa201YehCD6/OPTUZICDdT2/0Yhsh/zvIsOoyIrrPQaDSaXKSUJNImLuvEuSueRTJtcuGvX+KtLcNjucXqBFWK7E7a6bKMhdPpYKOcgqOEsYjjKdYrIONZdDsjuoJbo9FocrE7s+aGoUbrWWwbjvHQO9t5fv3AmO2vJizPYrvsoNWnsqFcDkGPnIJjuCf/2GSUsCxjLFw+MFx0OaLas9BoNJpc7JkPdhjK63KMOnXWjvPvds0j2Ks+39WFw/KQXA6DDXIqzuENYGb3k06ECZuu4rRZACHA106nESG8qwL3wHoY3LBr71EH6moshBD/JoR4WwjxlhBipRDCK4ToFEI8LIRYY9135Bx/uRBirRDiXSHEKTnry4UQb1rPXSeEEKU/UaPR7C6SqXzPwusafTZUMKaMxW7PpgpuJyG84G7JLLkcBhvlZAwzkQlTAcQjIVVjUcqzAPC20y7GIHX2jxfDL06AwZ5de58xpm7GQggxHfgKsEJKuRBwAGcDlwGPSinnAY9aPyOEONh6fgFwKnCDEMJhvd1PgQuBedbt1HrtW6PR1EbSVF5EbursaD2DUFxN3NvtaaehbQw7O2m2QlAATodgQFrGI5oNiyWjISKVjIWvnVbGIHV2sAcifXDbWRDbzRpOBeodhnICPiGEE/ADW4GPALdaz98KfNR6/BHgdillXEq5HlgLHCaEmAa0Sin/LqWUwK9zXqPRaMaJTBgqo1kYozYWGc9it4ehtjNgdNFitRcHcBkGwzSpH3JO1jIZJSbdtHhdhe+i8LbTQqj4dyClutVCOqV0lNnHwM7V8PrtI/k2daVuxkJKuQW4BtgI9ALDUsqHgClSyl7rmF5gsvWS6cCmnLfYbK1Ntx4XrhchhLhQCPGSEOKlnTt3juXX0Wg0BZQMQ+2isRgPzaKPDppyjYVDEJCWsYgOZdaNVJQoHtzOMqdNXzvNMpTvWZgm3HkO3HCk0iKqEdoO0oSFHwd3c22v2U3UMwzVgfIW5gD7AE1CiM9UekmJNVlhvXhRyhullCuklCsmTZo00i1rNA3Fc+v6ueWZxjlZFGJPk3M6sqmzozUWtsC92zWL0HZ2yHaac4yF02EwjNUoMJZvLCKVjIW3naZ0KP87PHsdrLoPBtfDzR+EbW9W3k9gq7pvnQ5tM2B4U+XjdyP1DEOdBKyXUu6UUiaBPwAfALZboSWs+x3W8ZuBfXNePwMVttpsPS5c12j2aG5/YSM/fnzteG+jLHYYyu3Iz4aStYZccgiNRxgqHoJEiG1mG83eMp5FThjKkY4Sle6KnoXXDBFLKP2FbW/Co/8DB38ELvobONyw8p8hUiE9OLBF3bfuA2377jXGYiNwhBDCb2UvnQisAu4DzrWOORe413p8H3C2EMIjhJiDErJfsEJVQSHEEdb7nJPzGo1mj6UvlCBd2Ca7gSjSLKwBSPHUyNNngzF1gq2rZ5FK5P9s1VhsSbfleRYuh0EQHxKRF4ZypGPE8OCp4FkYSBzJkPp50wsg03Dy/8Kk+fBPv1Gpuvd8QYWnSmGl8ipjMUMNXGoQ6qlZPA/cBbwCvGl91o3AVcAHhRBrgA9aPyOlfBu4E3gHeBD4opTS/su5GLgJJXq/DzxQr31rNI1CXyieKXxrRIrqLKz7Sif8/lC8ZNFaMF5nz2LtI/Dd6dD/fnYtOghAb8JfYCwEEoOkqyUbhpISZ7pKGMpq+eFOBdQsjHhArfu71f2M5XDKlbDmryo0VYrAFnB6VcvzthkQ6VcNDN9/DN74/ai//lhQ12woKeUVUsoDpZQLpZSftTKd+qWUJ0op51n3AznHXyml3E9KeYCU8oGc9Zes99hPSvklORo/V6OZYDS+Z2EL3NnUWah8wv/nXzzPNX99r2i9rmEoKVU4KJ3Ir12IBwEImN4CgVudFpOu1mwYKp3EkGmi0pMJuxWR00wwnjIhFgDhUNXdNodeAB1zlJZR6jQW2Aot01SRX/tMtTa8BZ64Cp68alRff6zQFdwaTQOSNiUD4YnhWdgnT4+zehhqZyjOhv5w0fqYFOUlo/CjJbD6/vz1d++H3tfV49y6BctYhPDR4s0XuAHizpZsGCoZUS/HXT4MldemPKXe39uqTvw2hgOO/CJseRk2Plf8HoFeNqc7+M973iTetI9aG+xR+kciUv13UEe0sdBoGpDBSAJT0uCeRb5mYZ9E46nyJ/xEyqQvnChaD41FGCq0Q2UdvXhTdi02DI/9L/i7sj9nNqO0hRDe/DCU1fYj7swJQ1nGIiG8lG0gkeNZRJNpFYbytBYft+TT4OuEZ68vfi6whTWxFm57fiMX3ac0FdY9rj4/UWxkdyfaWGg0DUh/SJ1Q06YcVXbR7iBh1VnYqbP2HO54hf5QiZRJfyhetD4mqbPWyZ91T0C4X4Vvfnka9L0H//h/6rlSnoX0lQxDxZ05YShrlkXS4S3/+TmeRTSRVmGoUsbC7YcFH4P1T+WvSwnBXnrNTuZPaebZnW5MHLDqz9YewrUX99UBZ7knhBDXU6aeAUBK+ZW67Eij0dCXc0JNmzJzQm4kisNQ6r5cYZ3d0tw2hLlksqGSaaSU5a/eK2Gd/JFpeP02eMXKPvr0XTD3H+APF5Y0FmF8eRXc9u865miGYL5nkapkLIo8CysMVYrmKZAIqopth/XZkX5IJ9gs2jl8Thcdfjc7t3cyZdga72qmlO7i9JR+zzpTybN4CXgZ8ALLgDXWbQmwZzZs12gahFxj0ai6RVEYymWHoUp7FnYRXzSZLsqICsZSmdD+aFJv1Qstz8JwwsPfhP41cNZvYb/jlW7gbctmKAHEg5iGiwSugjoLK6srNwyVsI1Fmb5QAO4mTOHMjlaND5f2LEDtBfL3Y9VYbEi20+x1cvyBk9mQ6sx/XaVQVJ29jrLGQkp5q5TyVlS9w/FSyuullNej6iWW1HVXGs1eTl/O1Xej6haFqbPVBO5Eznqud5FImcRTZmb40KhDUQnLszjgH1XLjOO/AXOPyz7vbSvSLJJOVXxXKgwVNVogFYNkLONZpCt5FkKQ9rTlexaeltLH2sYidz8BVWOxOd1Bi9fJ8QdMZqtUWktMWJ9bzlgMrIPvzoD1fyu/v12kFs1iHyD3GzdbaxqNpk5MDM8iP3XW66ochso1Frnfz9Yruq251pFqInc6BTvfUyfGnHkTZkwZi97l/58qgDv60vzXFRqLeJCEQxmL3DCUwxAIARGjWS3EhnPmb1fwLADT05avWZQLQ2WMRbboz/YstslOWjxO5k9pZtg9FYAXUvPUMckyGVFv/F5pNu/Ur165FmNxFfCqEOIWIcQtqCK779RtRxqNhr5gvmbRiIw0ddY2LpDvWdg1FpNb1dVzVc/iV6fCTw6FW09XxWr2+wTUiffRjRIOPgOMgtNbkbEIkTBUD6hczwKUdxHOGIuhzEnadFU2Fvg6aCNENJ4qnw1l7wXy9xNWzU/7aKPF60IIQcu0/QF4zjxIHWOL+IW8/Qd1v+7xyvvbBaoaCynlr4DDgXus25FWeEqj0dSJ/pz00lS51hDjzEhTZ/PCUOGsMQxasywmtyjPomLn2XQKtrwCs45WP4ez3aWTMRX/D8oyoaIiYxEgavgRAvxuR96hLkPkGIvhrLGo4lng76JbBIjHI0qQHkkYKhYg7fSTxpFJ5Z3zD+dxZfN/4ph5mDqmVK3F9ndUO/NJB0L/WhiqTz+pWlNnHcBOYBCYL4Q4ti670Wg0QHE2VCNiewp29pCnSupsIp01ArmajF2QN8kyFhVrLYJbVbbTfsdbb5qN4ZvRIHHpIpouc1rztBZpFlGhWn0UZl85HQZhkdOm3ApDSWdlY2G0TKFLBDAj1udUDUPlG6+USxkXW3Bfuv90vvG1r9PVaQndpcJQb/8BhAGnXa1+rpN3UTZ11kYI8T3gLOBtwP4rkMBTZV+k0Wh2ib5gHCFUgksq3ZjGwvYUXEZB6mwZzyJeRuC2w1CTLM2iYhjKnk09ZYH1psHMU2YsQAhveWNTQrMI05FXkGfjchgEbKk2NpQxSqarqfzeAEfLFDoJYNqV35628nuBImNhC+65FeUAhketm/FQ/hX+phfhpV+pYUlzjlWtQt5/HJadU3Gfo6GqsUBNpTtASllcSaPRaMYcKSV94QTdzR52BuMN7FmYOA2BYVU8Z8JQ5TyLMmEoW+Ce3GoJ3JWMxZCqOTC75gMG6WgQe26djIcIy0rGol1dmacS4HRDPETI8JYxFoKgyGlTbnkWhqtCNhRgNE/CECbesBUKKheGcjcrb6AgDBV3qNBXiyd/Gp/hUevJWIhMlcWaR2Dl2apD7WnfU+nBc/8B3vur6mpbqNnsIrW82zqgzBxBjUYz1gTjKRIpk6mW4Nuo2VApU2b0CgAhBG6nUVPqbG6YzS7Im9yivm9FzWJoAyBYFWklJL1s3p4zETMRIoyPWLkK8lsJGQ8AACAASURBVMLahniQgOnLq7GwcTkMQnlhqIiakudyFB2bR7MautYSsjygcmEowygOi8UDxBylPQuHW62nollPitV/UtXgFz0Jky0BfP6pMGNFfpbVGFGLZxEBXhNCPApk/oV1BbdGUx/sEM2UVi9vbhnGbNh2H2YmbdbG6yw/h9suyuvwu/LCUHZ78po0i6GN0DqdoYSgEy/pWDY7SCRC1cNQoE7Qvg5Ihgm4PCU9C6dDEDUd4PJnsqFiePCU6zhr06SmRLdFLWNRLhvK3k9BWCwiVKfZwuwsp095KOlYTp1FPKTan/s6smsLPqpudaAWY3GfddNoNLsB+6p7ihWWaVTNIpk2i2Y7eFyOqp7FtDYfO3JSg0OxFE5D0OFXAYyqmkX7TIKxJGHpzRbiAUYyRFj6iFc1FkOZFNShdJkwlGGQSpvWCX0IzDRRPJkq9bI0K2PREbOGFpULQ9n7KQhDhX0+PE6j6Pfq9nhJSYN0jkajiv6aK+9nDKlqLHSarEazexm00mbtIrVG1ixcBVfaHqdRNXV2n3Yvq7epAUGGIQjGUjR7nTXNw2BoA8w5lkA0RQgvLTl1B85kmBDTavAsAhlhfDDtKbqKB3A5hcr28rarMJThIIq7/CwLmyYVhupOWMaiXBjK3k9BGCro9dPiLY76e1xOIngw4zmeRSJU2XMZY2rJhpoHfBc4GNUnCgAp5dw67kuj2Wuxr8ztuHXj1lkUNzj0VNIs0lnPwpRw49/W8dDb22jzuWj2OHE7DAxRwbNIJdRwoPaZBGJJItJLW04qqTMVISS95V+fG4ay+kj1p8qEoQxD1ZH42tXxLp8afFRulkXmM9pJ4aArZbUXrxaGGlinHqeTkIwQkP4ivQJUdXwEL85cYxEPQOuMyvsZQ2oJQ/0KuAL4IXA88C9A47XA1Gj2EOyTrd+t/ns2qmeRKOFZeF2OsmEg+3tNa1fXnFc9sDrz3EHTWhFC4Hc7y3sGw5sACe2zCPalCOPDmcoKue50uDaBOzac8Sz6ky6mlRS4hQr/Nber4UP+LsLSXd1YGAYBRzud6X5wNalhR+Xwtmc9C2s/w6a3pLHwuRxEpIeWRIFmUSnMNcbUkg3lk1I+Cggp5QYp5beAE+q7LY1m78UWiJs86kTTqNlQyZRZFJap6FnYYag2VdjW2eTmc0fPAbK9mbwuR3ljYaXNKs0iRRgPrrTlWZgmHjNCCG/m99cfitM7HM2+PtdYWFpH0PSVDkM5LM9ixnLYuQo5uI6o9GRamlQi5LAE50ohKMh6LZDJ0BpI+0p6Ol6XgwheNdPCZjdrFrUYi5gQwgDWCCG+JIT4GDC5zvvSaPZa7JNtU4N7FoWps6D6Q1WrszhoWiuTWjxc/fFDuPy0AzlmXjeLZqgTuc9tECsXRhqyMow6ZlkCtw+PaRkD6ySaW2fx7T+/w0W/eTn7eneTmomd41mEKX1ydjoMkqaEAz4EgAhsVZpFNc8CCLusautqeoK3TekO6ZTSUYDBMmExZSw8iNwK7kRI1WvsJmoJQ30V8ANfAb6N8irOreemNJq9GVsg9je6Z5EuTp31uAwGSoxNtY8HmNHh48VvnJRZ/83nDs889pXzLMJ9sPYRdbJv2YdgbCdhvHilZSwsDUKFodTrdwTjbOjPObnaMy1yNItgwfxtG7dDqGyoyQdBx2wY7CGCp7rADUTdnRCmeogot+7D8ix2Jr0lBW6fy0Gv9GLYnkU6qdqnN5LALaV80XoYQukVGo2mjtgx96xm0ZgCt6qzKNAsavAsKl2d+9zO4grune/Cz4+DVFS1sXA4CcZV6qyPuGpTbs/TzvEsQvEUw9EksWQ6M/IVb2u+ZyG9+EoU2mUEbiGUd/HcT4jJGlJngZi7K/tZlchN5bU8ix1JNweWFbg9GKl+tWCn0DZSGEoIsUIIcY8Q4hUhxBv2bXdsTqPZG4mn0nicBk6rjcbEqrOokDqbNhGCzPcqhc9lFHsWax5ShuLCJ+GM6wHVfDBkJWeasWB2nrYlcEspM21EdubUdGQ8i0Q2DOVzlzAWtsANcOA/AtSWOgvEvcpYyApX/b/5ew//9+Q29UNsOONZ7Ei4S4ahPJZm4UhZnoWdMrwbBe5awlC/A/4deJNsI0GNRlMn4klTGQsrxNOomkUyLYtO/NUEbrfDKJ6vHdyuRqAaLpqcsDNSYCy2vgpt+8I+2QGdgWhSCb5AKBSgNZENQ4HSfcKWsdgRjLFvp5pbkQ1DZUeqekt4Fm6HkUn1Zd8jiE1eypub53BCDZpFyqdqLUx3C+Xk8Nc3D7O9T6rL9RxPJ2CWTp21s6GcKTvsZnkWu1GzqEXg3imlvE9Kud7KhtogpdxQ951pNHsp8ZQKm2Q8i4Y1FqWK8hxl233EU8WeCG/dDdcvh1s+BL88meNijxbXSWx9Nc9QgPIs4tbgonBgMOtZyOwApXDc0i4CJTyLeIiUU51ovSUynPI8C4eTdR+9j3vNozPNEiuR9inPwn7/UgSiSQbSVrvz2HAmKypIuV5VgihenOmYWrDnjTdSGAq4QghxkxDiU0KIM+1b3Xem0eylxJMmHpeBw+oa2qieRSJtZuZv23hdlYvy8k62j38H7jofJh8In74bgMmyL1+ziA6pwrVpxcbC16zCPOFQVrBOWy2+I8k04YTtWRQYi6gyLnY7cJ+7+DTodBh5xZB2aK2W1Flp9YdKuMqHiAKxJANpf/Y7xgNIw00cd0mBWwhBwvDikpZGk9EsGkjgRonaB6I6z+bOs/hDvTal0ezNxFMmHmfjexaptCxRZ1G5N1Tm+Cevhie/B0s+Ax++Fhwu8LbTJgP5nknv6+p+n6WZpVgyTSJt0tTSAVGIhYaR7iACcPlbYVi1TLH7L24PxLLvN3UxvPpb2PQcCctYlDIAboeR1yW3FnHeRjZPxZSCuKuNcuYiGEsRwDIWsWGIBUi5WyCSPw88l6TDp868iXC2J1aDhaEWSylXSCnPlVL+i3U7v+4702j2UmyB22HYmkVjSoUlU2edBmlTZtJkc0nYYaiNz8HjV8IhZyvB2mFdSfu7aDUD+QL31lfVfY6xsCfrtbW1AxALB0jF1MnT16wyjHbmtEDP8ywO+Sdw+mBoY2b+dinNwmmIPCNt6xe1GAtHyyT+OfkNdu5XPgATiCUJ40XaMy3iAVKW8SoVhgJIOSzjkgjneBaNVcH9nBDi4LrvRKPRACp11utyZIxFo3oWJTULlz2Hu4KxeOEXaoLc6T/MH9Dj76I5PUQ0mUbabsHWV6F9Fvg7M4fZ8y86OlSldCISIBUNYEqBv0mFZfpyDESeZ+Frh0UfV3t02GGoYmPhchp5Bs/2LGrRLHwuB8+ZB6uuuGVQBk8gPVnBPWFpHKUEboCUPdI1GWlYzeJo1DyLd6202Td16qxGUz+KPYvGNBYl6ywyc7iLRe5E2mSyMQzv3AtLP60G9+Ti76IpNYyUOcZm66t5XgVkPYvuzm4AUrEgqagaqdplDVDqtwoDXQ6RnzoLsOJzAMSEOvl6SxgAl6G6ztpGKz6CMJS/SvdcKSWBqDJ4af8k1fMqFiBmTckrlToLkM54FqGcbKjGSp09te670Gg0GeIpk2aPcwLUWciSYSgo71mclngEzCSsKBHJburCl1LtOaKJNF7DVC0+Fn8q7zDbWHRankU6GsQUQSL46GpyA1nPYmanPz8MBTB9GSz5NGsDc3AaAmeJ2gl7LW2qzroZzaKGOgvbYJYbDxtOpLHtf2zKMlw9D0HzFKLGVICSAjeA6bKNRURpFk4fOGo5hY8NVb+5lSa7L3CC9ThSy+s0Gs3osCuOG92zKJc6C6VHoyZSJsfFH4fZx0D3vOI39HfhTQ4BUmUyRQbUelN33mF2GKrJyoYy40HMWJCw9NJpGQtbs5g7qZmBcCJPrAbgozfwYsc/lqzeBjLfK2kZ6pEI3LZnUS6F2N4/QHjKCogOQP8aosKf9/pCssYivNubCEJtFdxXAF8HLreWXMBv67kpjWZvRmVDGTiteH4jahZSyjKNBMt7FvG0SUd6ACaXkUD9XTjMBE3ECERTEOlT6wXGImCdbFv8HqJ4MifPEF66rIFR9tjWuZOULrEzFOf9naG894klTTxljYUy1EkruWAkqbO2BlLOswhEU9nHk5arB2aKqNFUucLdZc0ET4Z3e3tyqM1D+BhwBqo1FlLKrVA2I0yj0ewiqoI761k04gxu+4q78Eo7o1mUMBbJZBqvjKjur6Xwq2K2DhFkKJqASH/euo0dhmrxuogJPyIRQiTUSNVMGMryLPbrVlffP35sDSd+/0ne3pqdTKc8uNKnwMIQ4Eg0izafCiP1h+Iln8/1LIJNszPfL2I0la5wt5D27832LHZj2izUZiwSUqk8EkAIUeZfOh8hxAFCiNdybgEhxFeFEJ1CiIeFEGus+46c11wuhFhriemn5Kwvt4T1tUKI60S536ZGswcQS6mTWCNrFnamUDnNolQIRqbjODDLh0/8yoPoJMhwJKk6zeas2wQsY9HscZJw+HAmQzQNr6VHTqWr2TYWyrOYY3kWK1/YBMDaHVnvIpZMlw9DOe0wlPqedupsLdlQfreT6e0+1uwIlXw+kGMsEmkJ+6quuxHRVNEYiVxjsZtHqkJtxuJOIcTPgXYhxOeBR4BfVHuRlPJdKeUSKeUSYDlK67gHuAx4VEo5D3jU+hkrPfdsYAFKVL9BCGH/S/4UuBCYZ9206K7ZY4lb4RHDEAjRmHUWWWNRe+qs026CVy6Dx7rC7hRBBiPJCp5FkmaPE4chSDj8zEquw5UK8oacm9EsBsJxDAGzOvMzrrYMZYch5XWjLcAWsu0OuiMRuAHmT2nm3W3Bks/lhqESaTNjLELCX9EYGe4G1yyklNcAdwF3AwcA35RSXj/CzzkReN8SyD8C3Gqt3wp81Hr8EeB2KWVcSrkeWAscJoSYBrRKKf9ueTi/znmNRrNHIaXMpM5CcXFYo2BfaRdmEtkx/VKps86UNVuibBhK1VJ0UBiG6sw7LBhL0WrVIqSdTcySmwF4w5xLq8+F0xCYEpo8TrqaPfhcDk5bOJV2v4stg1ljEa0QhrJDScNWims8pQoQjQodc3OZP6WFdTvDaiZGAblhqETKhFlHARAQrRWNkeGxNYtIY4ahhBDtwBBwJ/BtKeXDo/ics4GV1uMpUspeAOvenro3HdiU85rN1tp063Hhukazx5FMS0yZjf07DNGQ2VAZzSI3DPXeQ8x++AI8JEp6Fu605VmUDUMpD2KyI6TCUJF+1cvJkZ9KGowlM+mlaStDKGV4eE/OwO9yZEJLtvdx18VHcs0nFzO93VfgWZhlPYsOy0MZjKhwVl6rkhqYP6WFRNqkJ3f4koUdRrPfl30Phc/8gdd8R1SZ9eFim+xADqy3wlANInALIdxCiFuAHuDnqNBTjxDil0IId60fYB17BvD7aoeWWJMV1kt91oVCiJeEEC/t3Lmz1i1qNA1DNuvG9iyMhvQskqmCMNTb98Dtn6K55yHmi80lNQu3PS+73BWxtw0MJ9NcYYZszaJArwDlWdhVzm6fituvd+6H0+nC6TAyGU72bO0F+7TR5FE6Qq5nUSkM1eFXxijPWNSgV9jMn6JO5Gu2F4ei8jUL6/e0/4nE0kbFz/C6DF4256l2KQ0WhvovVJrsvlLKZZb2MBNVyPf/RvAZpwGvSCm3Wz9vt0JLWPc7rPXNqHoOmxnAVmt9Ron1IqSUN1p9rFZMmjRpBFvUaBqDeEFbicb1LHKMRWAr3H0BdM4FYK7YWtKzcKWtE3U5YyEE+LuY7Axnw1AFegXkG4vuLhWieiY6M2Mc7C6yTQWV0NM7lGdhV2VXMhbtfnU9PBRRJ/aE1dyxVvaf3IwQ8G4pYxEt8CzsxyWGSeXidTl42TwAMbxxt49UhcrG4kzg81LKzLe1Hl+CSqetlU+RDUEB3Ed2hve5wL0562cLITxCiDkoIfsFK1QVFEIcYWVBnZPzGo1mjyJrLNSJSWkWjShwqxOuy2HAe38FMwUfvwkpDOYavUXGQkqJx7Q8i0pXxP4uunMF7qZizyKQE4byNalmgq+l52aK2ewwVGH31untPiKJdMYAxJJmyVYfAO2+fM8inkqPyLPwuR3M7PSzZntxRlQglsxoIoWdbSuFurwuBy+aB2QXGkizMKWURQE3KWWIMmGgQoQQfuCD5Lczvwr4oBBijfXcVdb7vo3SRd4BHgS+KKW0fdmLgZtQovf7wAO1fL5GM9Gwwzd2VlGjexZup1BjT9tmwtRDkO2z2E/0Fo1WTaRNmoTV0K+cwA3g76JDBLKaRYG4DUp0tk+29nu9KedkjIU3E4bK9wRmdKheULZuEUulSzYRBCXct3idWc+iylV/KeZNbinpWQRjKbqtFN94obGo4lmskjMx7YaCDRSGkkKIDqsuIu9GjeNVpZQRKWWXlHI4Z61fSnmilHKedT+Q89yVUsr9pJQHSCkfyFl/SUq50HruS1I2YJWSRjMG2KmaeZ5FA9dZuEnCuidg/skgBKJ7HnNFL7Fk/ikikTJpwjYWFYRZfydtMshQJF5Ss0ilTYYiyUyKLPudwOB+H2GdnIbfrTwJb4FmYTO9XYnhmy3dolIYCqDd72IoR7OopcYil/lTmunpCxe1GglEk5lK80ROtlQ8beKuEOryugxSOIlNWaYWGkXgBtqAl8vcdAW3RlMHMgK37Vk4GtOzsE9yXTteUKmc81QNreiaxxzRSzyZzD8+ZdKEpVlUDEN105weJhkNqIaDBZrFoHWlbxffMfso2j59K13NvownkZsNlUuuZyGlrBiGAujwuzOfV3IkbBXmdDeRMiW9w9G89WAsmfEsRhSGsgxJpkXIbuw4CxW6zkopZ+/GfWg0GshckXsznkWDZkNZ3k5X7+Oq++mcY9QT3fvjEwm8kW2o+lqFHYYyhRPDUSGZ0t+FNxWgNT0ITni5z8Gk/ggzu5RXYGsIGc8CMAzB1Z9YhM9lexalBe52vwu/28HmwUgm/OMtE4ZSx7sznkV8hKmzAK1WqCyYkyoLKnW2zefCldPNFiCRU19TCjtk1j/1aKZyPbTvW/bYeqC7x2o0DUSRZ9GomoV1kmve+Rrsexi4rDh6l+om2xruyTveDkOlnH6V9VQOfxcGJnNFLwA/fn6QX/xtXeZpu0Fgpz/f4Jxw4BSO3E95IeU8CyFEJn3W1oa8FcI+HX5XxrMYaeosZIcYFRmLaJJWr6t4dGvVbCj1XF/nMviPdTApK3YPhBOc8P0nWNUbGNEeR4I2FhpNA1GYOtuo2VD2npyJQH7GktV6vCOyIe/4ZNo2FlVay00+EICTDDXXYkC25LXNGLCGGnU2l/dOvGWMBWTTZ+3BRJU0CxWGytUsak+dBWjxKM8iFM8ai3gqTTxl0upz4XYaeZpFtTBUXvv3AuF/VW+AdTvDZVuMjAXaWGg0DUSs4CRmiMb0LBJWGMpIhvOF1uYphPHRGduYd3w8pcJQaVcVYzHrKFKuFs5w/B2AAVpYvS2QqY0YKBGGKqScwA2wT7uP3uFYJtxn12SUot3vIhhLkUqbeS1YasWepR2K53SZzXTMdSpjMYJsKF+FORlbrQyvwiy0saSWdh/XCCEWVDtOo9HsOkWehaMxe0PZYSgjGcw3FkKwxTGDudE3YDDrXSRSJs1EMasZC4eL0KyTaLbSbAdkK4FYim3WHO0BKwzV4S9vLOyTarOn2BOY1OxhMJIgbF3tVw5DWYV50eSoUmftMFQoJwxlG4tWr2vExsI2gqWMRe+w+v2UKoYcK2r59quBG4UQzwshviCEaKvbbjSavZzCoryG1SzSJg7SGKlYUVbOU74TmJFYDz9aDG/cCagToV/EkdWMBWDOPw2AuHQSN5QWstoKrwyE47R6nUXdbnOxDUCzp3g8aVezGymzJ9dqqbOgqrhHkzprh8Fye0HZs7dbvE7cDoN4Ol+zqPy97PbvxQYh41mUeG6sqKXr7E1SyqNQldOzgTeEELcJIY6v2640mr2UeCYMlaNZNGidRTYVNt9YPNp6JpdMukUVzG1+CbCyoYgia6g69h10CnHpZIBWjp6n2vas7lXGoj+cyNQolH19pt1HsSGww1f2ybWaZgEwFEmMKnXW4zRwOUSeZmH3hVKahSNzcjdNSTIta/IsoqXCUBnPYhzDUADWXIkDrVsf8DpwqRDi9rrtTKPZC5konkUiLWkpYyw8LoOtsksJ31FVc2tnQ9ViLLzNbTwhl7FBTmHFrA6mtXl5d5vK8hmMJDJN/spRLhsKoKtJGZotGWNRuc5CfWZyxF1nQWVfNXuceWGo7QE1PW9yiydP4K5luFLFMFRGs6ifZ1G2zsJGCPEDVNfYR4HvSClfsJ76nhDi3brtTKPZC4kn0wiRnUDnNAwiqVSVV+1+kmmTZlG6yK7Z46SnLwztXZmZFAlL4BaVWn1YCCH4X/dXGQhF+cGUFg6Y2pIJQ/WHEszo8Fd8/cyuJvxuB5NbvEXP2cV8dvfZWsJQWwYjRJPpkoJ5NVq8rjzPYrulvUxp9eJxGCQsT8A2FpUMksMQuB1GkWchpcwRuMdXs3gLOERKeVGOobA5rA570mj2WuIpE6/TkZnD7DAEDRiFIpXOad9R4Fl0N3voDydU9XXE8izSJs3Eam5R4WtqJoyPeZObOXBqK+/vDJFMmwyEE5k52+U4dl43r37zg7SV8EDs1262Tq7lxqpC1ljc/+Y2pITD5xb3qapGs8eZN+yodzhKu9+F1+XIE7gzk/iqhLp8bgexRL6xCMRShK21UkOnxopaTOVrwIEFY6+HgQ25PZ80Gs2uE0umMwV5oDSLRhyrGk2maTNszyK/VXZnk5tgLEXa24Fjx2oAEok4HpEk5qnuWQC0+9y4nQYzO/0cOLWFZFry/s4Qg5FExRoLUJ5JuZqIdr8bIWrzLJo9TpyG4IWeAfxuBytmjcJYeJ15RXnbhuNMbVUej9tpMBQdobFwOYo8i605A53GNQwF3AAsA95ADSJaaD3uEkJ8QUr5UN12p9HsZcQLsm4cDSpwh+NpOl0qjbWwVbYd6om52mmyNAsZV1PyDG9tnsWsLj8SidNhcMgMlYD51Hs7SaZlUfX2SHAYgk6/m76Q0g4qaRZCCNqtYz+wX/eIBW6AVq8zk3kFKgw1tc0yFjkV3MkawlCgPItoQcZTbu+pUnrGWFHLt+8BllpDhZYDS1GhqZOAq+u2M41mLyReUCnsbNBGgtFEmg5H6TCULSKHjFY1/jMZQ8aV5mDUGIb69kcX8svzDgVUQ77uZjd/fVvNT6tUkFcLua+v5FlAdmLecQeMbphas8eZp1n0DsfyPIsRh6FcDqKJfA1r65D6d2jxOsddszjQmjUBgJTyHZTxWFfhNRqNZhSottm5noUxdsYiFYeH/l9GR9gVwokUnQ51dV5kLCzPIiCs8FR0ABLKs3DU6Fl4XY7MgCMhBCtmdfLKxkGgcquPWrD3J0Tl7CPIZkQdN2+UxiInDJVMm/SH40wpYSziI9AsSoWhnIZgRod/3I3Fe0KInwohjrNuN1hrHiBZ7cUajaZ2ijwLYwwruDc+B89eB+/eX/trUglY8wis+hOY2ZNUNJGm1TYWhWEo68p9AGs9MgBxNTHO6RtdW+0Vszuwp9jsShhK7U95Ph6ngajU1BDV1vyAKS2ZrrcjpdnjyqTO7gjGkZJsGKpE6mw1Y+F3O4gUCNy9wzGmtHrxux11rbOoRbM4FzVK9asozeJp4GsoQ6EL8zSaMaSwB9GY1lkMWe03dtaQ8W6a8OJN8PiVEBtSa5MOhI/fBFMXEU6kaBMxcPnBkX8asYvm+tK2sejHSCpjUatnUcihs7Pi8q6GoWzPoloICuBbH1lQNLxoJLR4nSSs3lLbLG0hV7OIF4ShPFU0C6/Lwc5gPG9t61CU6e0+nA5R1wruisbCKsb7k5TyJOD7JQ4pHjCr0WhGTSxp5qVzjmnX2cEedd/3XuXjpITb/gnWPgxzj4fDL4JkFP70VXj6h/CJXxJJpGk2SqfCqnYcgm0pK/MpOoBIqjCUGOUo0IP3ac1kAnXtYhjKNjaV0mZtWr2VCwCrkdsfatuwOsnbmoVnFJqFvyAMFUumWbMjxHHzJzEcTRa1Qx9LKu7MmoEd0f2gNJrdQ109C9tYVPMstr6qDMU/XA6fvQcOOA0WngnTl8HAegBlLIgUhaBAaQydTW62xq2iuEg/RsK6rqyhgrsULofB0pnteJxGTSf5StieTy2exa5iV5EHc5oh5gncaRMp5QgF7qyx+M3fNzAQTvBPK/bF4zTGPQwVA94UQjwMhO1FKeVX6rYrjWYvJZ40i+osxkyzsLvADm2AZAxcxRXOALxzLxhOOOzC/EFFnXPg7XsAiMRTNLuiZYvsupo8bIpZJ+PIIEYqoh6P0lgAnHPkbBbs01pVZ6iGramMtDHgaLCNRSieYnsghttpZIr93A4DKSFlypo1C29OnUUgluQnT6zl2PmTOHK/Lu54ceO411n8xbppNJo6Y1dw2zgMg/RY1VkM9oC3XWkQ/Wth6sLiY6SEVffB7GOKBuzQMQeigxAdIpJM43NGiwrybLqa3WyPpNTzkX6IW1e8owxDAZy6cCqnLpw66tfbZMJQFUaqjhV2RlcwlqJ3OMa0Nm/G2NmGIZEys55FFc3C7856Fr97biNDkST/cYqamOfJaUxYD6oaCynlrUIIHzBTSrl39oJa/RdIJ2HBR8d7J5o9nKIK7rGaZxEPQaQPFn4c3rob+t4tbSy2vw0D6+ADJQIHHbPV/WAPkXganzcCnuklP66ryU1Pfxh8HcpYJFyYCAzX6LKKxpJuW+Ae4eS70ZDRLOIptltZSzYljUUNYaiUKUmmTXr6wkxp9bBwulIJPC6D2DgPP/owquXHg9bPS4QQ99VtR43Ik9+DJ75b/bhEGHrfUJkkGs0oKEydHTPNws6E2v8kdNgEfwAAIABJREFUQMDOMiL3qvtAGHDg6cXPdc4BINW/jkTaxGtGyoehmj1qXra/C6IDOFJh4sJXef72bqKzydYsdl8YKhhLsi2QLciDHGORNjNzLWqpswClGQXjyYznAiqsNq7zLIBvoRoGDgFIKV8D5tRtR42GlND/vnLhZZX/tA9eBj8/Bq5dCH+/ofrxGk0B8VRxb6gxyYayxe1JB0DHLOVZlGLLKzBlATSXKEKzPItUn6rHdafDFYyFm0giTdqrPIupiQ0MOrtLHru7afe5MMTuCkMpYzEYSbLNCkPZ2CGnXM/C46i8p9zRqsFYKq8Nu8ep6ixknc47tRiLVImGgXvPWTC4TbUsSMXU43KEdsDrt8N+J0DXfvDXy+H352YqV8eTB9/alpdBoWlMpJTKs8iJWxtCYEo1HGeXsI1FxxzoPgD61pQ+Lh5UoaNSeFrA343Zn2MsygjWtogcc7fD8BYOSb3FmtYjduUbjBmGIehq9uBzjbzl+Eix53A/v66fRNpk0YxsYqntRcRHEIbyW8YimkgTiqcyxgiUZ2Fagnk9qKlFuRDinwGHEGKeEOJ64Nm67KYR6V+bfTy4vvxxL94M6QScdjWccx988Nvwzn3w6P/Uf48V2DQQ4Qu/fZlfPlNh75qGIJpMIyX4c64WnYYK26RLXS2ufwp+tETdAzx9rWrnUYrBHiU2+zqUd7HzXXj/seLj4sGyojWgQlGDPbhJ4jCTFbOhACKOdgjvwEOSDZ1Hl3/f3cx3P7aIzx9b/wCJx+nA7TB4em0fAIflFBfa4caRahZghaFiBcbClTU+9aAWY/FlYAEQB1YCAVQ1995BnrHoKX1MMqaqXeefCt3zVFz2qK/AoZ+DF26EbW+Wfl0qof5z1pEdQZXb/fA72+v6OZpdJ2xlDOUO2XFYQ5CKdItVf4LfnKkuYN77q1p74w6VjFGKwQ3QPkv9bR52IXTtr17/ym/yj0sEK6e3dszBObyh7EhVG7twLmio50PSy0DX8vLvu5s56eApHDi1glEcQ5q9TiKJNLO6/EzO0Sw8OZpFIp3GYQgcRmVNx+dWfxvRZJpQQRjKrhup10yLWmZwR6SU35BSHmp1nv2GlDJW7XV7DP1rweFRot9Amavzra+qTJNl5+Svn/Bf6kruL18rrV/c9yW4ei7c84Xy772L7AyqNtKvbRrKGA5NYxK2upM2ufMruKFEaOHx76oLk+75sGOVuvDoe0+ltpZicL3SKgDa94XPP6pe+8Yd+cfFg5UHFHXMxhXeSqewLnKqeBaDqOefNhfR5B//TKjxwL76L5yHUZgNVcvYVtuzsMNQzZ58gRvG0bMQQswXQtwohHhICPGYfavLbhqR/vfVVVjbjPJhKLt3Tsu0/HVfB3zgy7DpOQjvzH+ubw28cacSE9+5D351mkpZHOvth7N9ZB5btWPM318zdoSt1tN5noWh/osW1VoENsPso2H6ctjxDvSvATOl/hYLBfHQTvX3Nm1xds3dpLS16FB2TUqVYlvJWHTOQUiT+WKz+rmKZ9FvqucfM5dk4vd7G/bV/2Fz8rWgQmNhj9KthG0sQvFUCc3C8izGMQz1e+BV4L+Af8+57R30r1X/qTpml7/6t0NJpWK9bfuq+9z/lKB67Di98Om74IJHVPvoWz8yJu2jc+kPKc9iWptXh6IaHDsM1VxNs0iEITasLk4mHwTBXuh5Rj0nTYgH8t947cOAhPmn5K/7OvI9kVQczGTlwrkOFedfbLyvfi4TsvK7HXhdBm+7FjJ8wCd5MH1Y3oltb8L+91wxu8CzsLOh0mkSaRN3DXUfdjaUPbypUOAG6tbyo9ZsqJ9KKV+QUr5s3+qym0YjnVLeRNf+6j9JOc0iZiWLeUsYC2+7dUyOsRjaqDKnlp8HTd0w5WD4xM0wvDErVo4R/aE4bT4XpyyYytNr++o6SUuza9iehd+dX2cB5KfPBnrVfes+MHmBevzWXdnnC0NR7z2oDMvUQ/LXC41FpYsem32WknC18nHH3yoeK4RQLT8STaz9wNUEaMqrCdibaPW56GpyM7c7f6RsrmdROCGxHLax2GF1ns1LnbUE7lidai1qMRZ/EkJcIoSYJoTotG912U2jMbRBufZd+6sskEhfaUE6XiF+67VS5XI9i00vgEzD0s9k16YtUfeBLeX3s/ZRePOu8s/bh+0IctvzGwHoCyfoanZz2JxO4imTNdt1o+BGxdYsSnoWuZpFcKu6b5mmLjQANj2ffT7PW0jA2sdg3snFBXG+dkiGlUcBStyGygK3y8u6KacwSVgXSBVCVl3NbvpDiUwn1NzvtTfxlRPm8cOzlhT1tCpMna1lbKvfCkPZbcrzi/LGWeBGzbP4d1S67MvW7aW67KbR6LdcbduzgNKhqHgADJcKKxXisz2LnFIV+3FTTpGSrwNcTTC8ufx+nrwa7v6cuq9QeLPyhU385z1vEk2k6Q/F6W7yMH+KOgG8t72+2Vea0WMbC3+eZmF5FrmaRa5n0TIte0HSNlPd5xqLjc8qIzD/1OIP9HXmH1/poieHtybnVHdXCFl1NbnpD8czxqJ1Lw1DLZrRxrHzi4scC4vyahK43baxUMkqzSXDUOPkWUgp55S4za3LbhqNLZZN7No/ry9OEXYGSalWBqXCUHZMOdeFF0KJ6JWMRWArOH1qIM1fv1HWYAyElU6xeTBCf0h5FrO6mnA5BO/t0MaiUcloFu4cz6JU6myuZyFENhQ1+yh1n2ss1j4KDjfMPa74A+3iuxEaix73gawxp1c9trPJw0CuZ7GXGoty5KfO1uZZqOl+ZcJQ4yVwCyH+I+fxJwue+05ddtNIpFMqB32/E6Cpy+qLI0prCrFAab0CSoehYgHVAtrlyz+2bXp5Y2GaSsg8/EI4/Avw3E9U6m0Jg2Ebi40DEfqtMJTLYTC3u5m1OgzVsITjKQxM/ANvqmZ/iXAmGyovdTbQqy407Kv6yQep+1kljMXwJmifqbKfCikyFtbfRpXOsOFkmt+K06FrnvKGy9Dd7KYvnCAYU9OX91bNohyZMFTSJFmjsRBC4HM52BFQxqK1ZFHe7g9DnZ3z+PKC50r4tMUIIdqFEHcJIVYLIVYJIY60NI+HhRBrrPuOnOMvF0KsFUK8K4Q4JWd9uRDiTeu568SuNrSvhXfvV1dwh35e/extgxX/oorvCg1Gpdx0p1uNnszzLMp4IpU8i0ifylRpnQGnXqVScl/9LWx7o+jQwYgyFuv7wgxGEpmc93lTmrVn0cDss+0x/u75Mq6bjoe7zofrl7PPNpWlnudZBLaoEJTNjBXq4mPOsern3AuTcB/4y/RkKutZVC5WiybSPOA+Gb78EhjlTyFdzW4SKZNtgdj/3955h8l51An6rc65e/KMNMrJSpacA+C4DkQDNnEBe5cDlni7C0u4u11YA3fssQsLR7glmXCEJRgwwcZgbLBxQrItB9mysjTS5NQ51/1R9XWaTqNgBdf7PPN0z9fVX3/VoX71ywhRnT9iqC4k2K4ZClT4rBUNVamteUo+i+feDCUa3K/3fyM+B9whpTwD2AQ8DXwYuEtKuQq4S/+PEGIdSkCtRwmjL+m2rgBfBt4OrNJ/bQmroyH9wFeQ4cHqcMOrPwGdy+Gn76yu+ZSJgrtJM0Grh0DV+AZhtokxlRFeS1SbHkLa9LBRK3t1zGJWuOy2oVmkLJdkXt0X5OBUimT2+LVeNBwh936G63d+kGkRhuu/Dm+6FdxBNjz6MaAmGio2XJ3Tc+br4D1/Vkl3rkC1ZpGYqPaNVTJHWFjm0eZmqES2UJUL0ghrk7J/MknA7TjqpkWnG3N8Fm02Y/K67CVNs1401InwWcgG9+v9PwchRAi4BPg6gJQyK6WcAa4DvqWHfQuwmkRcB/xASpmRUu4FdgHnCyEGgJCU8gGpyil+u+I5x4WJ4QN4Dt7Lne5rkKLiLXL54ap/VglRI0+WjzczQ4FycteaoeqNDw+q23oRUSVhoXeUVv7GzME5Qy3N4pH9ahGw2khaTu5dY8YUddLx7G846FnDOzyfho03wMorYfU1OHNqt1+oNUNVahY2u9rEwNxw2GQbwsLK7Wmz9Wkqm68K721Ep96k7JtIHHUv69MRh92GTZRDZ9sWFjoiSmlrJ0eexSYhRFQIEQPO1Pet/ze2ce7lwDhwixDiUSHE14QQfqBPSjkMoG979fiFQOXKN6SPLdT3a4/PQQjxdiHEFiHElvHx8XpD2mJyVF3GrUMh/vXOmlLOXavU7WzFpWaizXdjnkh1NFRDzUILi3qmqJJTUy8S3g71o56tFhbpXIGkrjA7MhPHRpHugGWGUtf4rPFbnHzEhjnsGMTjroiocwWwF9LYKJZ9FsUCxEfnVguw8EbKwqJYVIKgkRnKHVTmqyozlKjv36ggkSm0JSy6tWZxcDr5vA2bbYXVh7tdBzeU83ACLge2ilpSJ8zBLaW0SylDUsqglNKh71v/t7NNcABnA1+WUp6F6t/94Sbj6+mossnxetf8FV2/6tyenjr1+NskGVM7rZ7uHr54924OTiXJ5ovcdMvDfPFRtWtn5kD5CY0WfwtPuFqzaDQ+pGVgI81C2CGgZasQSruo0Sws53bQ7eCbzn/hZsctpdILSzp9uOw2djYKn73332DPHxrPo5b99x/zjPPnJVJCfJQJ0YHfXbEI60XbR7qsWcTHVI5OqJGwqNAs0jNqbCPNQojq8VapjxbmomQ2j8/VhhlKf+9yBfm8zd5uhctuK5mh3G36LKyCgbXRZZawOV6Jt8ezVdQQMCSltLKFfowSHqPatIS+HasYv6ji+YPAYX18sM7x40Ymphb26y48A4DfPzPGg3smuWfHOJ+++yAxW4iiJSykbF18zVujWTQyQ1nCop5mER2GYL8yOVhEFqms7wosYbFxYYjNtt1cYHumtMNz2G0s7/HXz7UoFuGeT8HWbzaeRyXT+1U9q++8EnKp9p5jqE96BvJpRmVHtS9Am4P8pMt5FrUaZi2Vi39ClcVuqFnUjm/1PdYks+1pFlava8AIiwa4HPZ5JeVBWbOofU/tNoHTLk5obagjQko5AhwUQqzRh64EtgO3oRL90Lc/1/dvA14vhHALIZahHNkPa1NVTAhxoY6CekvFc44L2YQSFqsWL2RZt5/fPzPGndtH8DrtvO+KlezPd5IY26cG51Iqy7uZz6Kug7vOj9LpAX/vHNPSw3uneOLp7RRrTQ9NNIsLBwRBkWKZGCbkLO80Fnf6GJ6t40BPTqp+HO0WM3zqVnU7vA1+9X7TFfBo0E21RgqRKht0SViICs2ilJDXhmaR1MLC39X4tauERQtzqkYJi9aLv8dpL5mfAsZnUReP00Yym5+XGcpKzKtn2nM77CckGupY8F7gu0KIx4HNwP8EPgVcJYTYCVyl/0dK+RTwQ5RAuQN4t5TSWuXeCXwN5fTeDdx+PC86n1JaQDDcxWVrenhgzyR3PDnKZWt6OH9ZF4dkN/ao3v23k8jkCasfYrFQoYk0EC51wmfv3TmONz1G1FljWossUkKoogSJ5dw+N6wiWxyiiJgsd0XrCriY1AKlCsv0NbUXpGTr/mn+9TcNWm+CygMYPA8u/RA89l2lZYxubzze0BgtLIYKYXwNzFClaKhSoENdt1158Zdy/ppFNt7SuQ3KDFVlLmuCZYoymkV9lnX72TUWn1fobNkMNVcAux22E1pI8IiRUj6mfQhnSilfKaWcllJOSimvlFKu0rdTFeM/KaVcIaVcI6W8veL4FinlBv3Ye+TxajJrvZ72L9i9Ya44o5dsvshEPMM16/vxu+0ckt24Eof0wm+FGzYJna0s+dFKE6kjLIamU/SLKUZkzQ6xTkSUpVms9UyWx409Xbrb4XMxncjO7dNrCYvMLCSn+MZ9e/nC3buIPvzduR3VJnaqhk4brofLPgIv/7zqvPaDNzZ+DwyN0cLiUC5UvVvUwsJPpqxZJMYB0Tx3ophTod1WWfwan8WjB6a5+Rfb1XfgCMxQiWyh7f7VlikqaBzcdVm3IMSzo7F5RUM1MkOBJSxOTc3ilESmo6RxgUMV4PO57DhsgsvP6CXgdnBIduPIJ9WPrJ3Y9MqSH/VKfVRSR1hMTo4TEGn25WqeE9G1gGarhYVNQDitdqB57KrfgabT7yJflERTNbkW0bIbqDixiwf2TAISz13/Ax74UvXYJ28FBKx7pXKGnnMjXPRuVaE3YyKt5k1cCYv92VC1ecdtmaFS5Wgoa0FvlAxXmTuR1BuGGsFy6yOH+Maf9vLU4aiqDzUPYZHXCWT+NsxQUM61MJpFfdYNhMgVJIWinHfobD0B7HHajbB4LrFnoyRtalfndti54ZxBXnXWQsJeJ363gyGpf3wzB5SzGlrnWYCKiGqVJevrhFyyKjEvP612/dvjNSaCkmZRdnJPJbJEfC5sM/tJOTtIhlZUaRaWWWAqWWOKqojAGt63nalElh5mcGWmqp3zoJo59W+stpt365Diyja0hvaIjSDdQWYKLgJVZqiyg7ukWbRqe1opLBITSuN1uKqG7JlQAv2320fV+Gxct/ht0fgISOpIm3Yc3FBOCDWlPuqzfkF5HWjfZ6H9QHWEhcthO6FVZ5932HNxMrbyD/Lm6zbw6deoLmN+rVkASgNo12cBSrNoJVw8FSYrIFco4kioneejM97q5KxAnyoSV6FZTCezSvWf3oe3dwWhxWdWaRYdPi0sKjroAUqzCPSDsDG6T40/z3OofN2VZJPlRcmi6xgLi2IRnvm1almbr+NjOZ2IjVD09wFUaxaWz0JkKjSLePPaTVWaxURd5/aecVV94M7toxUm0pm2NItkxhIW7WkKlhnK5FnUZ1l3AI/OvJ5PuQ+oL4DdRrN4bnHn42Qd9X+QfpedwyVhcbC1WQnKAiA1o3wCzcZbP3YtLEZm0/QJZU7Yl4uwf7KizIjNRtY/wOjBsgN7Mp6l06eEBR1LVJG5mQMloWaZBaySICWih1WxxPAismM7Wdrl45rusfJ1V5JLzE3csgot1hMWQ1uVT2M+bP8p/OAN8JXL4LPrj3lTqHkzvE2FFh8PYiPkfEpY1PdZpClYDu7MPDWLGhNUIpNneDZNT9DN08NRJor6c0xOtdZaoFQqpn0HtzFDNcNuE6zpV2tBO82PoCIpr6HPwmgWzxnuQpy8s/6PxmG3kXSGydncyrGcbsNnUengbjH+wcPqx1hMKjvywekkHaiFfkqGlJ25gr25Tg7v28mhGZXrMJ3M0uWzKa2nY6nq8Q0w9gxQLsEwVRsRpYvTFTuX448f4OKV3awX+wGQ6Znq0Nhcam7FXKdXRWdN7GQOv/yvcOvb6s63LlLCA19UJSxu+IZ6/779Snjk282fU++1jxXbfgD3/K/jk4QYGybrVZFuvrpmqFQ5zyI7H81ico5ze++E2mz89QtUf5atY1ZI7iHVkrWFZnFwWn3PQt72zEqWGcqUJ2/MugElLI6Fz8I4uJ9Dsvkifpmk2ERTCLidTDv7VUJcO5U6PTWqPjQ0Q92vhcW+Q8rhfGg6RVCkkMJG1u5l+3C1sBgqdtInJvmPP6hGTVOJHMtcOnO3Y6lqpSlscMeHYeag0jqgOnxWSqVZhBYw5VrIIka4aHkXC9Jq8RWFbHXiXT1hAarvRz3NIh1VO3OrmVQrDj4Mh7bChe9SEVdv+z0sexH86gMw0cDMte0H8IVzy/6Z2Gh1scdG/Paj8KWL4eGv1i/gaGEFABxrn4zO3k65VWZ+VVKe3Ym0u/GLDEVZYYZyNduYWMJiSmsW1Wao3ePKX3HFGb0s7/bz8Ih+wPJ7tRAWP9pykLDXyUXLm+RuVLBpMMLqvgCreltHWT1fWbdgfsLC0zQayn5C26o+r5hJZQmKZNPF3+92MGnvVZpFJqpKkNub7JycXtVJL9U6GmrHjPpIntmn/BBD0ylCIgnuICt752oWw1kvIZL84M8HGY2mmU5mWWLTIZORJapHxg23KDPQVy7Dmx7D67QzXSksUtOQT0NoIXuLfUREgvM74nhj+xmWuptapd8im6jfx6BrFemRHXzvwf3Vxy1BYyXyteLBLyoBu1mH4rqD8Kr/UEmLt71X+TMqkRIe1BFbo0+p26/9BdzdRtuVZ36pBMCvPwA/unHuuTX5WZ0Md6yFhc7eTriUBlAbZSSdfp1nUeHgbqZZOL1qk7D77rpFBHePJ7AJWNLloy/kYSSrhb5VvbiJsJiIZ/jNUyO8+uyFpVj/Vizt9nPn311KT9Dd1vjnIxu0sGjXD+RrUO4DVOVZY4Z6jphJ5giRxOZtnDfhdzk46FgE48+oWPYW9f9VDZ5ItYO7zo+yWJQ8NaU+kv2WZjGTotuZRbjDbFgQ4slDs6UciWQ2z1jWQ0CkEcUcH/rJ4xSKkgVFvV20uvutfyW89U5lwrj9g3T6XVVmqFvv0RVZQgvYllQ7xt7ttyCQbLPrmpGVfosGmkU8uBSPTPHMzmerH8gl1e2TP23+PlnsvRfWvrzaLxLsh2v+p2oTuuPX1eOH/lzu6zG1FxKTSuuriAKrSy6lMtZf+Ldw7b/As3fAPfUFTGJC7bwL48fY1BUbBSDq1MKixhcgXQH8IlUObGgjYomz3gT77lX5PDU+iz3jcQY7fCq72uPgQD6itI+n9GfT5Nw/2TpEriB54/mL5zFBQys2L4rwf990Dpetaa+e3doFIdYvCLGmf+5n5XbYTtkM7lOOmVgCn8jg8EUajgm4HWy1b1a78V2/ayuRCU+kHDrrClTXeNIcnk0xllNVR+Mzk4xF0wxNJ+lxZsAd5MxFEaYSWYa03fjwTIooPgA+cvkC7tmhNIqe4qgyPVVm+fatg0s/CE/fxkucW6vMUIf2q93ymOjitpkljDoWIvROfaf/LDXI0iyKBShkwOmjUJRVyX27Cv0AeGP7yq8rpRIWvi4Ye6rkO2lKNqFCiGtZ+3J1O1OjuTz8FSWwfd0wtRusjPWasilzGN+h7PS9a+GCdyA3/yX88dNzzWVS4suobOjMyDwd9a2IKY1l1qGE9Jz8BZcfP5lqn0WrLOvNb1JFJwH81QvQ7vEEK3qUEA64Hcxkgc1/WXpP796b4ot319eefrx1iPOWdpSqFxuODUIIrt3QX6oa24qFES+/et+L6A165jzmdphoqOeM2KxyLDv9jYWF321ni1inwlaTk81zLCysYoKZ2YaayO7xBFmc5G0ewiLBPTvGGZpO0WFPgSfE5kF1TduG1MJ9cDrFrFQ//JvO7uB9V6wEICJSKly31jR28fugayU3ZH9WpVm4kkoTueOAnccnBD+94D9VJ7511zHhU+csaRZaS/jDvgTrP3oH332onOPxSEItTOHkvvJr5jNqQd74WmWKe/g/mr9PFcJoDq4gIKq1nEIenvqZagDUu1Yt9BNas5k5OLdmVSEPP3uXEhSW5tG7HoTgsc4XA7Bvb432kJzEiWoNesRmqEZFB3T29rRuGDmnqZDLX646m8+q+l0t2p4SGig37aoInS0WJXsn4izvUc8PuB3E03nVAVJz194k335g35xTJrN5do7FedGqI6/mbDj+eIwZ6rnDKk/uDXY0HON3O5jKOmHxRepAO5pFaKFaaNJR8s4AH7vtKeKZ6ixqqymR8EUYcGf4t9/uYHg2TUikwK3UTpfdxuNDKqx2aLqsWZCe4e+uWs3v338pfe5M/WuyO6F/Ix1ytkpY+NKj5KWN//OQOu+6Jf2qK+Brv10dhw8l/8Nvd0aREn79xHDpPA+Mu0hJF13pikq4lgmqYymc/RbV13xqD9z2Prijtltv+fx1hYXNpoRgbVHGYg66Vqjoqak9ZWGRT5XrI1nMHlC1rB7+iso/sbtLjYMe10P3HqjWSKR2bh8s9uCO7qv2a2STc68TlAZ572fgsxvhnzvgE33wp8/NFRp6LlOoDcSckFS3H7/QPotSc6I2vm8XvENtZrpWlg4dnk2RzhVZoYWF3+0gkSmo+a+4AoDJnJvRaGZON8WdugfKaqNVnNQYzeI5JB1XP15fE2ERcDvUQr/ySnVAawq/f2aUm255mH0TdaJwlr5QmUVGHicqvXzz/n187d7qCq+7x+NEfE5s3giXLHLQ4XNRKEr8MgmeEC6HjXULQjx2UF3j0HSSpE3/eNOzCCFY3hNApGONa1V5IviL8SphEciNM0GY8aTakZw5WH6u3Ved92Et/oO9Xbz5wiVs2TdNKltASsljh2JMEsKbr/Rv6MXU6YUXvV/5b756JTzyLdXPvKII4pzx9ajtOlgZjda5XDl1h7aWH68p4V7yGe24XTnDe1aXNLDHJ1Ufh9GxkaqnRMeV8HiwuBZ7MaM6JYISfJ9eMTfDHeDX/wB3/bPKP3nh38OKy+G3/6Qc9JXo658puJRrq8ZxLFzBcp5Faa6ti/2x/DL4b4fLfitg34R6b5d2K0Ec9DjIFopqJ3rJP8CiC5QPA9UKtRKrrH09O7nh5MHtUP0xjkf5PCMsasgmlBnK1dQM5SCRycOKamFx19Nj3LNjnJd/4T7u31Wzo112qbqd3kdKlxL52r17GY9l+NGWg+wcjbF7LM6KngDCEyEskvzyvS/kO289Hz/JkqawaTDMk4dmKRQlQ9MpPEFt269dQBuZxrwRPIUoqVyeVLZAJl/AW0iQdqjxy7r9RHzl8hAlc1yqWrNwegO8cFU32UKRh/dNMRrNMB7LkBR+PIUE+UKxajwuv4rMOuevIDXF/p7LlUllzz3V12cJi0bd2uqVewf1/nStUPcPPljOKK9tO2stuNFDKtGvd13poS2j6gc2MzlW9ZSZkX0A3F/UOSuTu5SGcP//Udcbr+nKmE3C9tuUJnXjbXDlP8IbfqAcz49+pzoMORMFh5d4VuB3ze1TLdx+1c+iSrNoQ1iA0iQr2D+lNjFLu8o+C1Cd71hyMbz1TmaySljVbnieHY3hdthY3FlH4zOcNBzPPtxGWNSQT6hdovDTVjKaAAAgAElEQVQ0joYKuB0kswWKPetgYLOqk4TKih4Ie+jyu/j4r2oicbpXqXIaQMqmfnCJbJ4r/u0e/uHHj3PTLX9mx2iMlT2BUmc9h93Gi1Z2I9LlznqbFkVIZgvsHo8zNJ0iGNE26aq2rbONTWOeCHZZwEeGyUSG6USOIEm8gQguh41Ng9XzDnjcRKWPQqmMtVrMnR4/FyzrwmW3cd/O8ZIfxeYNEyTFbErb+Gs1hatuhr+6nTdH30VU+ph+7BfV15c9Us0iWO5DLYtlrW+mRrOo1GSKuZKwGI9l2B8X5HBQTM0wU1E7KzF5iKIUPOHYoA5M7IJ998GEdnZnqsOZ2XmnynLfcH35mBCqpDtUm8Z0iY1Epn7Zb+EK4LP6WVhFGtvRLOpwYDKJy2GjP6Qco5Z/JJ4um5ws0+i+Gs1ix2iclb0B7LZqYWY4uXjtuYv43d9f0nbpkPlghEUNBd3LopnTurQjyxXgHX+AC/8GUHHoS7v8vOWipTw9HC35IAC1WCy7RD1P+HHaBX95wWK8TjsfvHYNo9E0M8kcK3r95TBbUBFXxVzpes7UTu5H9k9zaDpFuFM7HKuERZN+GdoHESbBVCLLRDxDQKSwecPcctN5vP/qNVXDgx4Hs9JPPq6ERVEnurm9AbwuO+cs6eAPz45z+xPD2G0Cb7CDoEgyndTCorT46x2p08PB4GYOzOb4Q/FMbLvurPYBlHwWbWoW2qyUdwWhY1n5+OB56j2ojYiyhIUW3JawUMmOgqwzRIR4ydQHkJ85xCQhBhavJIlXaS5WXkflOS2eulVFIS15YfVxK4w1WUdYZPP1K7m6AxWahXqd6bybrfunuH/XBNtr8m6asW8ywaIOb6lvs/U9jmXUZyWlVBozdTSLkRhrjL/ipKc74GZlb7CqN/exwgiLGoS16DbpT+GvVN8rmExk6Qq4eOmZAwgBv3y8pvurFhYx6SXocfLx6zbw0H+7knddtpJ/uEYt0mf0h/SCqK+jJkN8ebef5T1+vnTPbibiGfo6O8DmmLuANtEsAMJCCYupRJYgSeyeEC9Y2c2iGjND0ONkFj8FXeYilVAC0ONTu9sXrurm2dE4P3vsMK85ZxCHL0yQZHlnnqsRFsBDe9W5phZeQbgwzTOPVtR9yulFSmsWKa1FlahpUVvUwuJn22Pg8pXDhbtX1e0kWNICNr1OhRdrrfCpw+qcrkAXYZGoEhb2+AhTti4Wd/nZzjLV+GnHr8umxUrNIhOHZ++EddfNjUazEuQSk9Xj3UFmkjmC9UpouAJ4RI5iIVfSLF7/rSe5/ssP8MavPcRLPn8vX/1je90N908mWdJVFsKBmu9xJl8sJf/tq6hBNpvKMRJNs9r4K57XGGFRg8i2LjlumQtqo5km4hm6A276Qh7OX9rJL7YdrnY0aWExI/0EPco+bdmo337Jcm5918W8cGW3jviJqh13ujrj22YTfPTl6zkwpR3Nnf5q4WI1ZGriswAIac1iMpEhKFI4fPWFo6VZWA2hkgl1PV6/Ov8N5wxy/dmDfO9tF/Cp68/E7g1XaxZaWOTsHv7ljmc4MJnkoT2TRHxOXv26mwAY33Zn+QVLPg4fqWyBN339IV72+fvIWjZYbaKzsAISHh3TgtsyRXWtVP0+5mgW+v285IPwjntLZdafOhxlUacXp7+DAXe6Slj40mMk3D0siHh5c/oDZN74c3jZZ5VJDao1iwMPqCgsKyekkiaaxZ7xOMu762hT2ndjyyZLr5Oz+7jlpvP4/tsu5KLlXXzxnl3EM3nimXyV+awSKSUHppIs6SoLbSsDOK41i0TF97lSWOzUzu3VfUdm/jKcHhhhUcMVS70UHN45zsFKyjuy8o8rky8QS+fp0iWZX75pAbvHEzxxqMI81LEEbvgGv/dcNadksxCCsxd3KPXRGwH0op+Zaxa7dHUPV69TVUoXdni1cNHj8mmVudvIDKV9MZZmMRlXmoU7UN+hb2kWlsaVSqrdrd+vFo6+kId/e+0mLl6hFkKXP0KQFNNWCXS9+P9hb5wv37Ob9//oMR7aO8V5SzsJdvSRxI0tVbHT1maugt3Le773CFv3T5PKFUrCEU9E5WHo81r90p+e0kJ58Fzo26gW2cii+j4Lm1M93r+hdHj74SjrB8Lg7aDHkeLpihpc4cIEOV8/A2EPKTwMdZwL5/51uZ9IpbCwCg1aj1Vi5TzU+CzyTj+HZ9OlZLkqtLBw5BMlB7fdE+LyM3q5aEUXH3rxGcwkc3zstqe46jN/4O3f3jr3HMBEPEsyW2BJheYYKG16lKC1NIxl3f6q8NkdJWFhNIvnM0ZY1LDYn8PexLkNlWaosrCwSn536xo4L904QIfPyV9/889sq9ilsuF6hvKR5iWbK/tfNChUePN1G3jHJcvZNBip3m23qoKrzVB9zjQHppJMxxJ4RK5h9JelWdgy6vzppLqeQLD++d2BDpyiQCyur0Mv/t97dBK3w8af901zYCrJBctUFFdcBLBlKgSqFgJ/3JfgrmfGuOGcQaBcAK+qkRSQS86Sk3aenshTLEq44h/hbXepMeFFSuDWOsTdQeVD0hyeSbF3IsHGwTB4IwRknBmtGUXjcTqJYgsNMBBWprHhmbQ1WX3OSjNUk/ffE1EmwyrNIkpMqvNa+Q9V6MgnWz5ZMkPZPOVxmxdFuHxNDz/eOsTwbLpUfbiWAzoSqtoMpTZEloPb0pSthjxW+OzjB2cJuB0sjDQIOjA8LzDCopZ0ExOOxtIK4nWEhaVZdPhd/PidF+N12XnT1x4ilS37N6LpXPPOYZUNkBos/v1hDx95yVpVqbLSjm8tVo0Enl5slwdz7BiJkYwqx7VooImEPE6i+HBq81w2pRasQLC+cLGETipWHWq75VCaD117BucvVULiQl21NGUL4MhVLLbabPWTJ6YYCHv4x5cqB3RJWFRW8EUFJMTxksoVGY6mVRkVhy5aV6ftrCUssvliyU9x6yMqb+IVmxaAtwNfPkomXySdKzByaJ962c7B0mJ5eFYvyA63Suqr1CyaVSEWund2jWYxnVfRSSt7GwsLkY1DNkYWJ15vtV/pn16+nr96wVJefdZCphuYoawci8V1zFDWpieRtYSF+u7sn0wQz+T51RPDXL2ub05Yr+H5hREWtWSijU04mpJmUZHlOqHNLlazF1A7xb/7i9XEMnlGouXy17F0vrlmUdWGtY22rZVmqFY9wd1hQLDEl+XZ0RgpHeXU6PyWZmEvZiCXJpdWO9RIA83CCjnOWufVDmuby8drzh3kM6/bxEdefEaphn/aEcJdR1j8fk+cG84ZJOxz0hdys3tM29BrNItiurwzr4o+A2WGgmont44U++b9e3np5+/jd9tH+dHWIS5c3qmc+54I7kIcG0Vi6TzJSdUt0N25kL6w+mxLmoX1vqWj1ee3OcsCqxZ/d7k3tnqjGM+6sNtE1a6/hDZD5dMxyMRJCe+c786ybj8fffl6VvQGSGYLpOu01dw/lcQmYLCjrB1Y1Utjmfqaxdb90/zs0UPEM3nedNGS+vMxPG8wwqKWNjQLf42tFyrMUIHqfsdWW8nKNqaxdK5u45IS9dqwNu2XUVECo9V4mw08IQZcGaaTOcYmdEJZA+ES8DiIohex9CyFTJKMdBL2zy1iVvm6uURZs8hj56LVAwQ9TgY7fLzj0hWl0L6sM4S3ULEz16G2KeniNeeoxX55d6DUN7pWsyATJa5LnuyuEBbDsymiLt0jvNJvoT/fXz2hsrTf+/1H2T+ZLL2W1Q8iRILZVI5cVFWFdUdUobfugJthrVmMxzJEix5krWZRY+aqwtdV1izyGShkGU47WdLpq9/PQJu6Cuk4ZOMk8BJqoJVaLXMtE1olByYTDIS9VcXqbDZRrg9FWcPoD3t4+aYFfO2+vfz773aybiDEWYsaJ6kanh8YYVFLG5pFoE4y00RcCYPuQPWOsraNqZSSeCbfvhmq3eZK6VkdCdW8uZI1vtuhFrzE7EzT8zvtNlJ2q6TIDIVMgjQuHI2SfvTrFiw/QTZJClfDfgYFVxi/rNAIcknSuLhgeXfJZLKi18/usbiKLKvsOgjYMjFieHHYRMlU9fDeKa741z9w7uceIyvcZCb3lc+fiZK2+dh2cIZXnbWQgpQE3A5evFHnXVh5KCJBNJ2jGFXZ3P4OJXiWdPlKzu+v3ruH/Ql7KSJLnb9FH2t/d9lnoT+roaStVNxvDtoMJTNxyMRJSE/DftadfvWdmtMFEaVZVEZClS7HbS+boTJWy1QHn77hTDYvijARz/CmC5cYE5QB0+uwljf+p4q/b4LXaccmah3cGTxOW6k/rkVtG9NktkBRtuhJbGkWlhmqVXMlT1iVzsilWpuhALwRwjZl1gmKZMvxOWcICup6ZDZJ2uahYQiAFjoypa6jmE2SlO7SrrcW6QkTkKo8iMNug1ySlHQrx71mRU+AaDrPRDxLj6faDGXPxYhJH2sHQuwai7N1/zQ33fIwCyIeNi/q4MCTXQQO76bfOlkmxiFULsZ7rljJNev7yeQL5cYzWrOIECeayuFNKGER6FJnuHJtL//7jh0cmknxm6dGuFz6Srke1vmbCnZfdznPQn9W++N2lYxZD6vsifZZRKWn4XcnUtIs5gqLQ9MpLl/TO+d4qc4ZZU054HLgcdr5+o3ncdtjh7j+nIVznmd4/mE0i1o6l1cVX6uHEKqOT62Du8vvnrMDsxzeVv+ImNZGmmoW7qDqR5CeVX+tmivV7fHdXLNw5WJEfE4CaGdtkwiwvLvCLJZLkhMNTFBQ1mgsh3g6Rkq6S+a4WoQ3QkikmE0oP0A+kyCBu6rHsxUltHs8Xm2iAxy5BAnhZ+1AkKeHo7z921voDbr5/tsu5APXrOaQ7MYZP1R+wUyMPVEbq3oDrOgJcO2Gfq7bXLEYViQtzqZy2FMTzEg/Ho+y9b94g9IwPve7Z9k/mSSGF5muiYZqpVlkZlW5cR3dNFv0qDIv9dDCQmQTFDNxYkVPw+9OyeRZR1jMpHJE/HOfF/A4S9/jsmZhL53vphcsa7vPguH0xgiLI6RUTFAzkcjO8VcAeJx2fC57SbOIpZU9uWkDeyHKfohmCXalF7EW0EqzVZMFyxNGpGZY3RdsS7OQ7vJu3pZPk7c3ERZaSNmzMaSU5NMJUrjpaCAs7LrJVHRWmWby6QRp6aoWFjpKaM94QkU7uUMlzcKVj5O1+0vaRzZf5Os3nUdvyEOn38Uh2Y0vWRYWMhNjd9TGNetLukY1Jc0iQTSdx5WeZEaUtZxl3X7O6A/ywy0qgiqGF5Gt9FlEKbgCfPyX27n+y/fzxbt3MRarcIhbPbGTk6XPKoavNMc5aDOUs5CimI4Rp5lmod6z6RqfRTpXIJsv1vV1BNz2KmHhcdoamxgNz2vMt+II8bvt1dFQscwcf4VFp9/FpPZpWJEnTc1QoHMnplvbwK2xUBYuTl/TpEKr9tTqvgBBS7NoookUrQUuMY69kKRgbxJv7wogEfhkgmS2QCGdIIWLzgZmKEdAhdImZpSwKGQSJHETqnh/BkIePE5bdfis1ixcxQRZR4CzFnfgtAv+/fWbS5qI22FnzN6HNzej8j3yGUQhQ1R6WdUoG9nKcBcJoqkcnuwkUXt1ufqXbFTaxYoeP3HpxZ4r+1yK6Sj3Hczw9fv2ksjk+fRvdvCSz93L/bsnmIxnSDr1uZITJWERl14WdTSo5mqzk7d58AslLBJybjSUhWXqm67xWUR1UcdwnXIilQ7ueCbf0B9iMBhhcYQoW29FNFQiQ1cdzQJUOG2tGSrUSlh0r4Lhx3Wdp1aaRUXPiTYc9FaL1zVasyg2C/UEHN4IGVwQH8FRSCMbVYQFsNnIOZQQmk5myz6LOiYQAI8WFslZZccvZJKkcFctbDabYGVvoJxV7dVJiLk0Tpmj4Axw/rJOnvjYNVy5tq/q/FG31iBmh8qLM966C2fpvQG6bEpYBHJTJJxzhYUQ8IbzFxPHizMfLzU1yiWjDCUc/OtrNnHH317Cb/72EkJeJ2/86kOc84nf8Z6f68isRFlYJPCUtIJ65J1+wiQQ2TgJGpuhnHYbQbdjTq5FVGuzoTpz9lf4LFTlWyMsDPUxwuIIqTRDSSmVz6KBZtHld80xQzX1WYCqIzW5U/WUno8ZqlkRQQuvKplxzeowZ/XZEZ5Q41BPIOh1MkEEGRvFWcwgHc0zeQuuICGRZCaZQ2bV4t/IZ+ENKWGRtvIysglS0j1nMT93SSePHJhWNaIszUIvtgXdOc7jnGtbT/gWqDszB0sO5bj01l04AXC4wOmnx5FSyZOFGdKurqohK3sD/Pp9L+Kmi5dScAawy4IqswLYc3FieHnBSvWcNf1Bfv7uF/DfX7KWl24cYH9aaxDJydL15J0BnE1MP/GuTVxqfxx7XvlImm00In7nHM1itolmEaxycDeofGswYITFEVMpLKKpPPmiLDmza+msEhbqOS3V/aUvUrep6fYd3Kl2fRxqfK8zzYULnIgWwiXocTAmwxSiI3hJI1zNG+BId5AgSaYSWUQ+RRpXw2gof1iVWM/FVU0lmUuRwjVnMb9gWSfpXFHV2ir1M1eLrWzSZjQfUOVCmNlf5SNolKsAgLeDLnuSZCJJkARZT/ecIWsHQjjsNoou/V6no5DP4ihmiEsvEW95vkGPk7ddspzrNi9gSuprTUyUaj3ZWny+0dWvZkBMYaNIQjbWLAA6fa45PotoqrE2a32PrZBuY4YyNMIIiyMk6HGUbMHj2h/RKJegy+9iMpFFSlmhWbT4UfZvLOdbtBIW1uOWg7sdzaJqfPPzBz1ORooRitERvCKLzd0gzFNj1w2QDs+ksOVTZG3eurt+AH9Y7cCLSaVZ2PJzzVAA5+taUg/tnSzXwmoju90W6ieHQ5X8aMcMBeCN0GlLIBMqYbHgmyssSljvdSZWWvxTNh8e59yfVl/IwwwBpLBBYhwyMYrYcHmbv59y9bWlLPUE3qbBERGfa44ZqplmEfA4yBclmXyRRKZQtwGTwQBGWBwxC8JeRmMZ8oViKdqlmYM7my+SyBaIp/MIQWt132ZXfbuhtabgcCmndmq6TR9HRRZ0G+ODHgfjMoJIjOIlg6OFsHD6OwiJBAenkzgLqaZmK5sWXDJlCYs0Kemes/PvCrhZ2Rvg4b1TZTOUDlm1N3l/OgNehmUncuZgaXxMegl5m5Vb6SAsEhTiur2qv6fhULtXv3YmWhJeBWegbhJbf9iDxEbGGS45uJPCR7iB1mURDAa5o6C67MWb5FmA+q7Nx2dhVRKIpfPGZ2FoihEWR8hgh5dCUTI8m2ZoSkUUNYposez1k/EM0bRS9dvqZKX7X7Rc/EH1bxje1rz/tkVlYlsb44MeJ2MygjM7S4AUTk9zYWHzhOmwpzk4lcJZTFc1PpqD00sWR2khdxRS5GyeuqUvLljWyZZ90xTcYcglKerkNru3cY5Ih9/FULGb4nTZDJVz+JvnDnjChGSc/KwSFrbg3GQ2C4f12plY6fyywefVHXBjExC3R0oO7qRoXL7DIux18pOi+i6M07xiccTnZDpRbYaa1Wapeq9TWUHZmKEMzTDC4ggZ1IJhaDrFgakkdptgIFI//8CKkppMZFURwXZ/kMsvBwQE+1oOZfllcPAhSE3NI4lvpnm/bk3Q42AM9Ry7kLi8LZrgeEIESXJoKoaTHMLdRFgIQcIWxJFV5UqcxTTFBprIBcu7iGfyDOfV62dHtgPg9DcWFp1+F0OyR2kWpez25iXo8XcTKUwSKio/iis80HCoVWVXVuS4NPIB2W2CnqCbYccgjD0NmRhx6WnsbNe4HXYes2/g2syn2Grb0FTQdfpcxDP5crMolGbhddrrCuDKCspGszA0wwiLI2RRp1rQhqaTHJhKsjDibRjR0qnrQ03Fs6qIYKtIKIue1fA398HaV7Qeu/xS1as7n27fDGVpFm2aoSx8/hY+EXcIn0wyMaVLcriaayIpexBnLgr5DDaKpZaqtZy9WF3Do5yhzvvMLwAa9uIA5S/aJ/twJEaUkxuweVu8P4svwleIcqltGwCeSGNhbTWNyibLwqLZ+ftDHnbYVqgot+hhorJ5dJNF2OvkGbmYoKe5ySrin1vyYzaVa2h2s4RFNJ0jkS0YYWFoyHEVFkKIfUKIJ4QQjwkhtuhjnUKI3wohdurbjorxHxFC7BJC7BBCXFNx/Bx9nl1CiM+Lk6Cq2UDYixBKs9g/lWRxZ+Pdc5e/XB9KFRGcxw+yf0PzBDuLxReBXS8k7Ti4HV6Y2tNWqG1Im6FKT/e11izsFCjqgnnOFg7crCOAJx8r9+tuIFz6Qx5sAp4tLIDQIM7JZ9TLBZprFg8WVU8MnvkVeRz4vM2juVhxJRLBVbatJKSbUKhJP/aQel+SsXIor6OJWaw35OHRnCr3LYcfY7boae5s11hjWm00rOTHyoioaCrf8DUsZ/l4TAVpBIyD29CA50KzuFxKuVlKea7+/8PAXVLKVcBd+n+EEOuA1wPrgWuBLwkhrG/ul4G3A6v037XPwXU3xeWw0R/yMDSd4uBUsqqpTC1zzFDzERZtX5AfFl2g7rfyWdjssPQFsON2kIWW4yM+Z5Vm0WgxL6E1lT6U09rlaS5ccs4wvkJZWNgahOY67DZ6gm5GohlYeSUAaekk4Gv83nf6XWyTK8g5AjC1h4TwtTT7EOhhKrwet8gzIcNNF/NASO11solyKK+riVmsP+ThgZQK5xXFvHa2z0dYNP/udPjmVp6dTeUa+kUsTWJU91sxmoWhESfCDHUd8C19/1vAKyuO/0BKmZFS7gV2AecLIQaAkJTyASmlBL5d8ZwTymCHl2dGokwlsk01C5/LgcdpYyqRIZbOEWjXDDVfll2qbttxiC+/HGZ1NnELzWKww8c/ve4SJFqha5bBDaUkwV6hzFDuFppIwR3GLxOlxkr2JtFWfSEPo7EMrLoKUDkTTfMO/C4K2BmKqGiiRKuwWc3kgHIoTxBuml0dCQZISyfZxCzFtNIs3E3MYv1hD3tSfmRQ+UHisrWDG+YhLOqYoaLpXMM5R/Rxq3GUcXAbGnG8hYUE7hRCbBVCvF0f65NSDgPoWyvUZCFQ0dKMIX1sob5fe3wOQoi3CyG2CCG2jI+PH8Np1Geww8d2XYKimbAA1ddiMn4cNQuANdeqHs+dy1qPXXFF+X4rhy/wsrMWI6wQ0mbRTVDWLITSLDy+1hnoYZEgrvt2O5pEW/WFPIzOpmHZJRSFg1iTWkmgFj+X3caOgFJso23u5FNL1PszRbjpAtrhcxHDSz41QyY+TUEK/E18Or06FyfVvQFQwqud67HGtFrMreTHqTk+i/qv0RVws2FhiF89Pgy0EdJteN5yvIXFC6SUZwMvBt4thLikydh6fgjZ5Pjcg1J+RUp5rpTy3J6exrHxx4rBDq9VEqilsOj0u3h6JKb7bx+nH2T/RvjwARjY1Hps71oI6LpJrXwcFgHt6G2lWehoqyV25bPwBZprFsIbIUSCmG7E5GwSbdUXcjMaS4MnzGjnOYzR0fT9FELQ4XfymOMsQJUDb2cnLxaezbDsZNg+0LTxT6ffRVyqMuW5pOoHHmmQyQ9KswCYCq0HVIJg05wPTbs+i1Ll2QozVDTVWLMAeNVZgyR0j3hjhjI04rgKCynlYX07BvwUOB8Y1aYl9K3OfGIIWFTx9EHgsD4+WOf4Caeyn3EznwXA5kURnh6OkitIehok7x0TWvkTLISAFZer+618HBZWCG8rzaJvPdicXO54EoBAoLkw8oW7sQvJoSEVreRuIiz6Qx5mkjnSuQJ3rPkE78u+p+Xi3+l3syvfQzGyjEkZassMFfJ5eFnmk3zf9+am48JeJ3G8kImTT0WJ4Wt6/r6QEhaHvKsBnSB4DM1QHqedgbCHZ0eVWalYlMQy+aYRVy/fNICV9mPMUIZGHDdhIYTwCyGC1n3gauBJ4DbgRj3sRuDn+v5twOuFEG4hxDKUI/thbaqKCSEu1FFQb6l4zgnFyrXo8Dlb/uA//soN/Pm//wU/fMdF/OUFS56Ly2vNqqvVbaCNPI7Kca2EhcsPiy9kSVFZFYPB5mau3h513tFDe4HmPo5evdiORTOMFUNM2ztx1+tdXUGX38VkMsfYK/4fN+fe3PZOfpIwHn9zrchuE6RtfmzZGIVUlLhs7hOxhMUux2ryNg8HZc8xjYYC2DQYYduQ0tJimTxS1s/etugNenjhKqWJm3IfhkYcT82iD7hPCLENeBj4lZTyDuBTwFVCiJ3AVfp/pJRPAT8EtgN3AO+WUlo1wN8JfA3l9N4N3H4cr7ttLM2ilQnKoifo5vxlnXhdJ8kPcv2r4F0PQteK9sa3a4aCstZCa2HhDiu3VXFiFwA+X2NNpF8vtiPRdClnpVUktVXIccqzmBG62lqcrR18pI2x044eutL7EOkZZYZq4hAPeRx4nXb2pv18/0V3cmfx3HlFQ7WTk7FpUYT9k0mmE9lS/bJWr3HjRUvoD3lKwsxgqOW46ZxSyj3AHOO5lHISuLLBcz4JfLLO8S3AhmN9jUeLlWuxqE1hcdIhhPJdtEvncrA5yxngzVhxBdx1M0DrjO9FF1BEcEH+zyDAG2iS1KZt/qPRdNvBAp1+F1PxbLlGUhu7c4fdht9lb0uwbPNdzNXRe/BNPsqTch2LmzxHCEFfyM1INI3PHQBhayujv10zFMCmRUo4bxuaKdUrazXnK9f2zekFYjBUYjK4jwKXw8Ybz1/My85sXA7itGLT6+FdDzTt112ifxN4VaXYlpqIr5OJwBksFKrWUzMfR1+wLCymEtm2Fv6BsIdYJs/+SRWa285OHuCVZy3k8jMa14WyONB1MWlc2GWudUVbYEmXn11jcdVcqc06YfMxQ21cGEYI2HZwtmmXPINhPhhhcZR88lUbuXbD80RY2J2qg1872GyqXhWobPEWpBaVA+Wama1CXpWzMjSd4rukGV0AAAleSURBVJH906VddDM2Dqox9+5U0VntLpyffNVGrttcN0q7ijWL+rm7oJTopPA1L1IIrF8QYtdYnIl4pu1r2bAwzI0XLeHiFV0txwY9Tlb2BNg2NFNRcdY4rg1HhxEWhuPHi/4erv6EEhwtCG1QzvactBMKNDbrKTOOhzueHCGRLXDJqtYh0tZO+0+7lLBoRxuZD5sWRbi9cD6AyhRvwfoFYfJFyZZ9021fi8dp55+v20CkRTnzymvadnCmaS8Lg2E+GGFhOH70b4SL39vW0I41LyKFmxRu/C0CAPpCHkaiaRw2wcUrmzQm0lg77elkDiHas/vPhzMHI/y+eBZR6WPW1druv26B8smMRNPHbce/aVGEyUSWh/aqyrntmt4MhkYYYWE4OXC4edZzJinhaRndZEXsnLOko+28gDMHlVO+7V4i8yDsddLX08Mlmc/yx3DrSjRLOn0lgXi8dvxXre2j0+/i1kcOYRMQMJnZhqPECAvDSUP6ipvZcubNLcf1h1SEz6Vr2s/S36x9G8faBFU+fwczBAn6W4ee2myCtQOh43o9/WEPt9x0Hj6XnaDHecwFpOH5h9luGE4aLjj/Yjj/4pbj+sPKYd6Ov8Ji0yKlWRyvnfzmxRF+8shQ2+dfvyDElv3Tx9U8tGlRhO+89YJSFJjBcDQYYWE45Xj1WQvp9DtZv6DNMiXAGf0hXHbbcfMRnDVPYbR+wfHVdCzOWdLBOUs6Wg80GFpghIXhlKPD7+JVZw22HliBy2HjktU9bWfbz5c1/UEWdXo5o7+9ooyWk9uEtBpOFYSUdQu4nvKce+65csuWLSf6MgzPI6SULZ3zFoWi5LO/fZY3XLCYhZE2yqcYDM8RQoitFc3qSphtjcFwjJhPt1+7TfCBa9Ycx6sxGI4tJhrKYDAYDC0xwsJgMBgMLTHCwmAwGAwtMcLCYDAYDC0xwsJgMBgMLTHCwmAwGAwtMcLCYDAYDC0xwsJgMBgMLTltM7iFEOPA/hN9HfOgG5g40RdxjDmd5nQ6zQVOr/mcTnOxOJFzWiKlnFOl87QVFqcaQogt9VLsT2VOpzmdTnOB02s+p9NcLE7GORkzlMFgMBhaYoSFwWAwGFpihMXJw1dO9AUcB06nOZ1Oc4HTaz6n01wsTro5GZ+FwWAwGFpiNAuDwWAwtMQIC4PBYDC0xAiLI0QIsUgIcbcQ4mkhxFNCiP+qj3cKIX4rhNipbzv08S49Pi6E+ELFeYJCiMcq/iaEEP/e4DXPEUI8IYTYJYT4vNDddoQQf6OPPyaEuE8Ise5Unk/F4zcIIaQQYl4hhCfTXIQQNwkhxivO8V/mM5eTcU76sdcKIbbra/neqToXIcRnK57/rBBiZj5zOUnntFif+1EhxONCiJccyZzmIKU0f0fwBwwAZ+v7QeBZYB3wv4EP6+MfBv5F3/cDLwT+BvhCk/NuBS5p8NjDwEWAAG4HXqyPhyrGvAK441SeT8U1/BF4EDj3VJ0LcFOzc56ic1oFPAp06P97T9W51Ix5L/CN0+Dz+QrwTn1/HbDvaL9/UkqjWRwpUsphKeUj+n4MeBpYCFwHfEsP+xbwSj0mIaW8D0g3OqcQYhXQC9xb57EBlFB4QKpvwbcrzh2tGOoH5h21cDLNR/Nx1A+t4flPobkcNSfZnN4GfFFKOa1fa+wUnkslbwC+P5+5nKRzkkBI3w8Dh49kTrUYYXEMEEIsBc4CHgL6pJTDoL5AqA+7Xd4A/Kf+8GtZCAxV/D+kj1nX8G4hxG7UAvu++Vx/LSd6PkKIs4BFUspfzvviazjRc9Fcr80BPxZCLJrHa9blJJjTamC1EOJPQogHhRDXzm8GZU6CuVjXsQRYBvx+Hq9Zl5NgTh8D3iSEGAJ+jdKYjhojLI4SIUQA+AnwtzU7/CPh9TTe2Yg6x0pfIinlF6WUK4APAf/jSC/gRM9HCGEDPgu8/yhf+4TPRd/+AlgqpTwT+B3lXeYRcZLMyYEyRV2GWtC+JoSIzPfFT5K5VD7/x1LKwtFcxEkypzcA35RSDgIvAb6jf1dHhREWR4EQwon6YnxXSnmrPjyqVURLVWxLRRdCbAIcUsqt+n97hZPrZtTOYbDiKYPUVy9/wBGaQE6S+QSBDcA9Qoh9wIXAbWL+Tu6TYS5IKSellBl9/KvAOfOZx8k4J/3Yz6WUOSnlXmAHSnicinOxaLYwt8VJNKe3Aj8EkFI+AHhQhQmPCiMsjhAdefB14Gkp5WcqHroNuFHfvxH4eZunrLKXSikLUsrN+u+ftAobE0JcqF/7Lda5tW3T4qXAzlN1PlLKWSllt5RyqZRyKcrB/Qop5ZZTbS76WgYqzvMKlC173pxMcwJ+Blyur6sbZZbac4rOBSHEGqADeKDdOZzkczoAXKmvay1KWIwf4dTKyGPgJX8+/qEiGSTwOPCY/nsJ0AXchVqw7wI6K56zD5gC4qidwbqKx/YAZ7R4zXOBJ4HdwBcoZ+B/DnhKX8PdwPpTeT41Y+5h/tFQJ81cgP+lP5tt+rNpep5TZE4C+AywHXgCeP2pOhf92MeAT51G68E64E/6O/cYcPXRzM36M+U+DAaDwdASY4YyGAwGQ0uMsDAYDAZDS4ywMBgMBkNLjLAwGAwGQ0uMsDAYDAZDS4ywMBiOAUKIgk6YekoIsU0I8fetsmaFEEuFEG98rq7RYDgajLAwGI4NKakSptYDV6Fi7D/a4jlLASMsDKcEJs/CYDgGCCHiUspAxf/LgT+jyiwsAb6DqggM8B4p5f1CiAeBtcBeVM2ozwOfQtVccqMqu/7HczYJg6EJRlgYDMeAWmGhj00DZwAxoCilTOvSLN+XUp4rhLgM+ICU8mV6/NtRvSE+IYRwo7JwXyNV/SWD4YTiONEXYDCcxliVQZ3AF4QQm4ECqpZSPa4GzhRC3KD/D6MK9BlhYTjhGGFhMBwHtBmqgKoy+lFgFNiE8hM2angjgPdKKX/znFykwTAPjIPbYDjGCCF6gP+LapcpURrCsJSyCLwZsOuhMVRJdovfAO/Upa4RQqwWQvgxGE4CjGZhMBwbvEKIx1AmpzzKoW2Vqv4S8BMhxGtQlWcT+vjjQF4IsQ34Jqp68FLgEV12epxj3J7VYDhSjIPbYDAYDC0xZiiDwWAwtMQIC4PBYDC0xAgLg8FgMLTECAuDwWAwtMQIC4PBYDC0xAgLg8FgMLTECAuDwWAwtOT/AwIcvnjm5HFRAAAAAElFTkSuQmCC",
+ "image/png": "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",
"text/plain": [
"<Figure size 432x288 with 1 Axes>"
]
@@ -2656,7 +2738,7 @@
},
{
"cell_type": "code",
- "execution_count": 2,
+ "execution_count": 1,
"metadata": {},
"outputs": [],
"source": [
@@ -2705,6 +2787,7 @@
"metadata": {},
"outputs": [],
"source": [
+ "import numpy as np\n",
"data, special_days = get_stalliion_data()\n",
"time_horizon = 6 # predict six months\n",
"# make time steps first column\n",
@@ -3135,12 +3218,12 @@
"output_type": "stream",
"text": [
"Missing timestamps detected. To avoid error with estimators, set estimator list to ['prophet']. \n",
- "[flaml.automl: 07-28 21:26:03] {2478} INFO - task = ts_forecast_panel\n",
- "[flaml.automl: 07-28 21:26:03] {2480} INFO - Data split method: time\n",
- "[flaml.automl: 07-28 21:26:03] {2483} INFO - Evaluation method: holdout\n",
- "[flaml.automl: 07-28 21:26:03] {2552} INFO - Minimizing error metric: mape\n",
- "[flaml.automl: 07-28 21:26:03] {2694} INFO - List of ML learners in AutoML Run: ['tft']\n",
- "[flaml.automl: 07-28 21:26:03] {2986} INFO - iteration 0, current learner tft\n",
+ "[flaml.automl: 08-13 01:07:10] {2540} INFO - task = ts_forecast_panel\n",
+ "[flaml.automl: 08-13 01:07:10] {2542} INFO - Data split method: time\n",
+ "[flaml.automl: 08-13 01:07:10] {2545} INFO - Evaluation method: holdout\n",
+ "[flaml.automl: 08-13 01:07:10] {2664} INFO - Minimizing error metric: mape\n",
+ "[flaml.automl: 08-13 01:07:10] {2806} INFO - List of ML learners in AutoML Run: ['tft']\n",
+ "[flaml.automl: 08-13 01:07:10] {3108} INFO - iteration 0, current learner tft\n",
"GPU available: False, used: False\n",
"TPU available: False, using: 0 TPU cores\n",
"IPU available: False, using: 0 IPUs\n",
@@ -3179,15 +3262,15 @@
"name": "stdout",
"output_type": "stream",
"text": [
- "Epoch 19: 100%|██████████| 129/129 [00:56<00:00, 2.27it/s, loss=45.9, v_num=2, train_loss_step=43.00, val_loss=65.20, train_loss_epoch=46.50]\n"
+ "Epoch 19: 100%|██████████| 129/129 [00:58<00:00, 2.20it/s, loss=49.6, v_num=0, train_loss_step=44.50, val_loss=62.50, train_loss_epoch=48.40]\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
- "[flaml.automl: 07-28 21:46:46] {3114} INFO - Estimated sufficient time budget=12424212s. Estimated necessary time budget=12424s.\n",
- "[flaml.automl: 07-28 21:46:46] {3161} INFO - at 1242.6s,\testimator tft's best error=1324290483134574.7500,\tbest estimator tft's best error=1324290483134574.7500\n",
+ "[flaml.automl: 08-13 01:28:51] {3241} INFO - Estimated sufficient time budget=13011770s. Estimated necessary time budget=13012s.\n",
+ "[flaml.automl: 08-13 01:28:51] {3288} INFO - at 1301.4s,\testimator tft's best error=1513915657450885.2500,\tbest estimator tft's best error=1513915657450885.2500\n",
"GPU available: False, used: False\n",
"TPU available: False, using: 0 TPU cores\n",
"IPU available: False, using: 0 IPUs\n",
@@ -3226,15 +3309,15 @@
"name": "stdout",
"output_type": "stream",
"text": [
- "Epoch 19: 100%|██████████| 145/145 [01:03<00:00, 2.28it/s, loss=45.2, v_num=3, train_loss_step=46.30, val_loss=67.60, train_loss_epoch=48.10]\n"
+ "Epoch 19: 100%|██████████| 145/145 [01:13<00:00, 1.98it/s, loss=48.3, v_num=1, train_loss_step=52.00, val_loss=71.40, train_loss_epoch=47.80]\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
- "[flaml.automl: 07-28 22:08:05] {3425} INFO - retrain tft for 1279.6s\n",
- "[flaml.automl: 07-28 22:08:05] {3432} INFO - retrained model: TemporalFusionTransformer(\n",
+ "[flaml.automl: 08-13 01:53:06] {3552} INFO - retrain tft for 1454.1s\n",
+ "[flaml.automl: 08-13 01:53:06] {3559} INFO - retrained model: TemporalFusionTransformer(\n",
" (loss): QuantileLoss()\n",
" (logging_metrics): ModuleList(\n",
" (0): SMAPE()\n",
@@ -3917,9 +4000,9 @@
" )\n",
" (output_layer): Linear(in_features=16, out_features=7, bias=True)\n",
")\n",
- "[flaml.automl: 07-28 22:08:05] {2725} INFO - fit succeeded\n",
- "[flaml.automl: 07-28 22:08:05] {2726} INFO - Time taken to find the best model: 1242.6435902118683\n",
- "[flaml.automl: 07-28 22:08:05] {2737} WARNING - Time taken to find the best model is 414% of the provided time budget and not all estimators' hyperparameter search converged. Consider increasing the time budget.\n"
+ "[flaml.automl: 08-13 01:53:06] {2837} INFO - fit succeeded\n",
+ "[flaml.automl: 08-13 01:53:06] {2838} INFO - Time taken to find the best model: 1301.3580441474915\n",
+ "[flaml.automl: 08-13 01:53:06] {2849} WARNING - Time taken to find the best model is 434% of the provided time budget and not all estimators' hyperparameter search converged. Consider increasing the time budget.\n"
]
}
],
@@ -3981,7 +4064,7 @@
},
{
"cell_type": "code",
- "execution_count": 9,
+ "execution_count": 8,
"metadata": {},
"outputs": [
{
@@ -4000,17 +4083,17 @@
"5248 0.090\n",
"6793 2.250\n",
"Name: volume, Length: 2100, dtype: float64\n",
- "Agency_01 SKU_01 2017-07-01 2017-07-01 77.331932\n",
- " 2017-08-01 2017-08-01 71.502121\n",
- " 2017-09-01 2017-09-01 88.353912\n",
- " 2017-10-01 2017-10-01 60.969868\n",
- " 2017-11-01 2017-11-01 60.205246\n",
- " ... \n",
- "Agency_60 SKU_23 2017-08-01 2017-08-01 1.713270\n",
- " 2017-09-01 2017-09-01 1.513947\n",
- " 2017-10-01 2017-10-01 0.993663\n",
- " 2017-11-01 2017-11-01 1.144696\n",
- " 2017-12-01 2017-12-01 1.989883\n",
+ "Agency_01 SKU_01 2017-07-01 61.183556\n",
+ " 2017-08-01 58.315655\n",
+ " 2017-09-01 68.994713\n",
+ " 2017-10-01 55.561470\n",
+ " 2017-11-01 63.187683\n",
+ " ... \n",
+ "Agency_60 SKU_23 2017-08-01 3.096925\n",
+ " 2017-09-01 2.331218\n",
+ " 2017-10-01 1.883857\n",
+ " 2017-11-01 3.282242\n",
+ " 2017-12-01 3.580535\n",
"Length: 2100, dtype: float32\n"
]
}
@@ -4024,15 +4107,15 @@
},
{
"cell_type": "code",
- "execution_count": 10,
+ "execution_count": 9,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
- "mape = 2743417592614313.0\n",
- "smape = 52.37\n"
+ "mape = 2670450710119213.0\n",
+ "smape = 52.3\n"
]
}
],
@@ -4526,7 +4609,7 @@
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
- "version": "3.8.1"
+ "version": ""
},
"vscode": {
"interpreter": {
diff --git a/website/docs/Examples/AutoML-Time series forecast.md b/website/docs/Examples/AutoML-Time series forecast.md
index 8f34efa9f..526eb3f78 100644
--- a/website/docs/Examples/AutoML-Time series forecast.md
+++ b/website/docs/Examples/AutoML-Time series forecast.md
@@ -426,7 +426,7 @@ plt.legend()
![png](images/CO2.png)
-### Multivariate Time Series (Forecasting with Exogeneous Variables)
+### Multivariate Time Series (Forecasting with Exogenous Variables)
```python
import pandas as pd
@@ -465,7 +465,7 @@ multi_df["temp_above_monthly_avg"] = multi_df.apply(
lambda x: above_monthly_avg(x["timeStamp"], x["temp"]), axis=1
)
-del multi_df["temp"], multi_df["month"] # remove temperature column to reduce redundancy
+del multi_df["month"] # remove temperature column to reduce redundancy
# split data into train and test
num_samples = multi_df.shape[0]
@@ -475,7 +475,7 @@ multi_train_df = multi_df[:split_idx]
multi_test_df = multi_df[split_idx:]
multi_X_test = multi_test_df[
- ["timeStamp", "precip", "temp_above_monthly_avg"]
+ ["timeStamp", "precip", "temp", "temp_above_monthly_avg"]
] # test dataframe must contain values for the regressors / multivariate variables
multi_y_test = multi_test_df["demand"]
@@ -503,46 +503,36 @@ print(automl.predict(multi_X_test))
#### Sample Output
```
-[flaml.automl: 02-28 21:32:26] {2458} INFO - iteration 15, current learner xgboost
-[flaml.automl: 02-28 21:32:26] {2620} INFO - at 6.2s, estimator xgboost's best error=0.0959, best estimator prophet's best error=0.0592
-[flaml.automl: 02-28 21:32:26] {2458} INFO - iteration 16, current learner extra_tree
-[flaml.automl: 02-28 21:32:26] {2620} INFO - at 6.2s, estimator extra_tree's best error=0.0961, best estimator prophet's best error=0.0592
-[flaml.automl: 02-28 21:32:26] {2458} INFO - iteration 17, current learner extra_tree
-[flaml.automl: 02-28 21:32:26] {2620} INFO - at 6.2s, estimator extra_tree's best error=0.0961, best estimator prophet's best error=0.0592
-[flaml.automl: 02-28 21:32:26] {2458} INFO - iteration 18, current learner xgboost
-[flaml.automl: 02-28 21:32:26] {2620} INFO - at 6.2s, estimator xgboost's best error=0.0959, best estimator prophet's best error=0.0592
-[flaml.automl: 02-28 21:32:26] {2458} INFO - iteration 19, current learner xgb_limitdepth
-[flaml.automl: 02-28 21:32:26] {2620} INFO - at 6.3s, estimator xgb_limitdepth's best error=0.0820, best estimator prophet's best error=0.0592
-[flaml.automl: 02-28 21:32:26] {2458} INFO - iteration 20, current learner xgboost
-[flaml.automl: 02-28 21:32:26] {2620} INFO - at 6.3s, estimator xgboost's best error=0.0834, best estimator prophet's best error=0.0592
-[flaml.automl: 02-28 21:32:26] {2458} INFO - iteration 21, current learner xgb_limitdepth
-[flaml.automl: 02-28 21:32:26] {2620} INFO - at 6.4s, estimator xgb_limitdepth's best error=0.0820, best estimator prophet's best error=0.0592
-[flaml.automl: 02-28 21:32:26] {2458} INFO - iteration 22, current learner lgbm
-[flaml.automl: 02-28 21:32:26] {2620} INFO - at 6.4s, estimator lgbm's best error=0.0925, best estimator prophet's best error=0.0592
-[flaml.automl: 02-28 21:32:26] {2458} INFO - iteration 23, current learner xgb_limitdepth
-[flaml.automl: 02-28 21:32:26] {2620} INFO - at 6.4s, estimator xgb_limitdepth's best error=0.0820, best estimator prophet's best error=0.0592
-[flaml.automl: 02-28 21:32:26] {2458} INFO - iteration 24, current learner extra_tree
-[flaml.automl: 02-28 21:32:26] {2620} INFO - at 6.5s, estimator extra_tree's best error=0.0922, best estimator prophet's best error=0.0592
-[flaml.automl: 02-28 21:32:26] {2458} INFO - iteration 25, current learner xgb_limitdepth
-[flaml.automl: 02-28 21:32:26] {2620} INFO - at 6.5s, estimator xgb_limitdepth's best error=0.0820, best estimator prophet's best error=0.0592
-[flaml.automl: 02-28 21:32:26] {2458} INFO - iteration 26, current learner rf
-[flaml.automl: 02-28 21:32:26] {2620} INFO - at 6.5s, estimator rf's best error=0.0862, best estimator prophet's best error=0.0592
-[flaml.automl: 02-28 21:32:26] {2458} INFO - iteration 27, current learner rf
-[flaml.automl: 02-28 21:32:26] {2620} INFO - at 6.6s, estimator rf's best error=0.0856, best estimator prophet's best error=0.0592
-[flaml.automl: 02-28 21:32:26] {2458} INFO - iteration 28, current learner xgb_limitdepth
-[flaml.automl: 02-28 21:32:26] {2620} INFO - at 6.6s, estimator xgb_limitdepth's best error=0.0820, best estimator prophet's best error=0.0592
-[flaml.automl: 02-28 21:32:27] {2458} INFO - iteration 29, current learner sarimax
-[flaml.automl: 02-28 21:32:28] {2620} INFO - at 7.9s, estimator sarimax's best error=0.5313, best estimator prophet's best error=0.0592
-[flaml.automl: 02-28 21:32:28] {2458} INFO - iteration 30, current learner xgboost
-[flaml.automl: 02-28 21:32:28] {2620} INFO - at 8.0s, estimator xgboost's best error=0.0834, best estimator prophet's best error=0.0592
-[flaml.automl: 02-28 21:32:28] {2458} INFO - iteration 31, current learner xgb_limitdepth
-[flaml.automl: 02-28 21:32:28] {2620} INFO - at 8.0s, estimator xgb_limitdepth's best error=0.0791, best estimator prophet's best error=0.0592
-[flaml.automl: 02-28 21:32:28] {2458} INFO - iteration 32, current learner arima
-[flaml.automl: 02-28 21:32:30] {2620} INFO - at 10.3s, estimator arima's best error=0.5998, best estimator prophet's best error=0.0592
-[flaml.automl: 02-28 21:32:32] {2850} INFO - retrain prophet for 2.2s
-[flaml.automl: 02-28 21:32:32] {2857} INFO - retrained model: <prophet.forecaster.Prophet object at 0x000001B1D3EE2B80>
-[flaml.automl: 02-28 21:32:32] {2234} INFO - fit succeeded
-[flaml.automl: 02-28 21:32:32] {2235} INFO - Time taken to find the best model: 4.351356506347656
+[flaml.automl: 08-13 01:03:11] {2540} INFO - task = ts_forecast
+[flaml.automl: 08-13 01:03:11] {2542} INFO - Data split method: time
+[flaml.automl: 08-13 01:03:11] {2545} INFO - Evaluation method: holdout
+[flaml.automl: 08-13 01:03:11] {2664} INFO - Minimizing error metric: mape
+[flaml.automl: 08-13 01:03:12] {2806} INFO - List of ML learners in AutoML Run: ['lgbm', 'rf', 'xgboost', 'extra_tree', 'xgb_limitdepth', 'prophet', 'arima', 'sarimax']
+[flaml.automl: 08-13 01:03:12] {3108} INFO - iteration 0, current learner lgbm
+[flaml.automl: 08-13 01:03:12] {3241} INFO - Estimated sufficient time budget=7681s. Estimated necessary time budget=8s.
+[flaml.automl: 08-13 01:03:12] {3288} INFO - at 0.8s, estimator lgbm's best error=0.0854, best estimator lgbm's best error=0.0854
+[flaml.automl: 08-13 01:03:12] {3108} INFO - iteration 1, current learner lgbm
+[flaml.automl: 08-13 01:03:12] {3288} INFO - at 0.9s, estimator lgbm's best error=0.0854, best estimator lgbm's best error=0.0854
+[flaml.automl: 08-13 01:03:12] {3108} INFO - iteration 2, current learner lgbm
+[flaml.automl: 08-13 01:03:12] {3288} INFO - at 0.9s, estimator lgbm's best error=0.0525, best estimator lgbm's best error=0.0525
+[flaml.automl: 08-13 01:03:12] {3108} INFO - iteration 3, current learner lgbm
+[flaml.automl: 08-13 01:03:12] {3288} INFO - at 0.9s, estimator lgbm's best error=0.0525, best estimator lgbm's best error=0.0525
+[flaml.automl: 08-13 01:03:12] {3108} INFO - iteration 4, current learner lgbm
+[flaml.automl: 08-13 01:03:12] {3288} INFO - at 1.0s, estimator lgbm's best error=0.0406, best estimator lgbm's best error=0.0406
+[flaml.automl: 08-13 01:03:12] {3108} INFO - iteration 5, current learner lgbm
+[flaml.automl: 08-13 01:03:12] {3288} INFO - at 1.0s, estimator lgbm's best error=0.0406, best estimator lgbm's best error=0.0406
+[flaml.automl: 08-13 01:03:12] {3108} INFO - iteration 6, current learner lgbm
+[flaml.automl: 08-13 01:03:12] {3288} INFO - at 1.0s, estimator lgbm's best error=0.0406, best estimator lgbm's best error=0.0406
+[flaml.automl: 08-13 01:03:12] {3108} INFO - iteration 7, current learner lgbm
+[flaml.automl: 08-13 01:03:13] {3288} INFO - at 1.1s, estimator lgbm's best error=0.0393, best estimator lgbm's best error=0.0393
+[flaml.automl: 08-13 01:03:13] {3108} INFO - iteration 8, current learner lgbm
+[flaml.automl: 08-13 01:03:13] {3288} INFO - at 1.1s, estimator lgbm's best error=0.0393, best estimator lgbm's best error=0.0393
+[flaml.automl: 08-13 01:03:13] {3108} INFO - iteration 9, current learner lgbm
+...
+ silent=True, subsample=1.0, subsample_for_bin=200000,
+ subsample_freq=0, verbose=-1)
+[flaml.automl: 08-13 01:03:22] {2837} INFO - fit succeeded
+[flaml.automl: 08-13 01:03:22] {2838} INFO - Time taken to find the best model: 3.4941744804382324
```
### Forecasting Discrete Variables