Update app.py

app.py

@@ -10,6 +10,9 @@ os.system('pip install graphviz')
 os.system('pip install python-pydot')
 os.system('pip install python-pydot-ng')
 os.system('pip install -U scikit-learn scipy matplotlib')
+os.system('pip install wandb --upgrade')
+os.system('pip install tensorboardX --upgrade')
+os.system('wandb login 5a0e81f39777351977ce52cf57ea09c4f48f3d93 --relogin')
 
 from collections import namedtuple
 import altair as alt
@@ -22,6 +25,14 @@ import graphviz
 from sklearn.metrics import mean_squared_error
 from sklearn.model_selection import train_test_split
 import matplotlib.pyplot
+%load_ext tensorboard
+import os
+import datetime
+from tensorboardX import SummaryWriter
+import wandb
+from wandb.xgboost import wandb_callback
+
+wandb.init(project="australian_rain", entity="epitech1")
 
 """
 # MLOPS
@@ -100,9 +111,36 @@ data_dmatrix = xgboost.DMatrix(data=x,label=y)
 
 X_train, X_test, y_train, y_test = train_test_split(x, y, test_size=0.2, random_state=123)
 
-xg_reg = xgboost.XGBRegressor(colsample_bytree = colsample_bytree_input, learning_rate = learning_rate_input, max_depth = max_depth_input, alpha = alpha_input, n_estimators = n_estimators_input)
-
-xg_reg.fit(X_train,y_train)
+class TensorBoardCallback(xgboost.callback.TrainingCallback):
+    def __init__(self, experiment: str = None, data_name: str = None):
+        self.experiment = experiment or "logs"
+        self.data_name = data_name or "test"
+        self.datetime_ = datetime.datetime.now().strftime("%Y%m%d-%H%M%S")
+        self.log_dir = f"runs/{self.experiment}/{self.datetime_}"
+        self.train_writer = SummaryWriter(log_dir=os.path.join(self.log_dir, "train/"))
+        if self.data_name:
+            self.test_writer = SummaryWriter(log_dir=os.path.join(self.log_dir, f"{self.data_name}/"))
+
+    def after_iteration(
+        self, model, epoch: int, evals_log: xgboost.callback.TrainingCallback.EvalsLog
+    ) -> bool:
+        if not evals_log:
+            return False
+
+        for data, metric in evals_log.items():
+            for metric_name, log in metric.items():
+                score = log[-1][0] if isinstance(log[-1], tuple) else log[-1]
+                if data == "train":
+                    self.train_writer.add_scalar(metric_name, score, epoch)
+                else:
+                    self.test_writer.add_scalar(metric_name, score, epoch)
+
+        return False
+        
+xg_reg = xgboost.XGBRegressor(colsample_bytree = colsample_bytree_input, learning_rate = learning_rate_input, max_depth = max_depth_input, alpha = alpha_input, n_estimators = n_estimators_input, eval_metric = ['rmse', 'error', 'logloss', 'map'],
+        callbacks=[TensorBoardCallback(experiment='exp_1', data_name='test')])
+
+xg_reg.fit(X_train,y_train, eval_set=[(X_train, y_train)])
 
 preds = xg_reg.predict(X_test)
 
@@ -119,6 +157,8 @@ st.write((cv_results["test-rmse-mean"]).tail(1))
 
 xg_reg = xgboost.train(params=params, dtrain=data_dmatrix, num_boost_round=10)
 
+os.system('tensorboard --logdir runs')
+
 #xgboost.plot_tree(xg_reg,num_trees=0)
 #matplotlib.pyplot.rcParams['figure.figsize'] = [200, 200]
 #matplotlib.pyplot.show()
@@ -126,3 +166,8 @@ xg_reg = xgboost.train(params=params, dtrain=data_dmatrix, num_boost_round=10)
 #xgboost.plot_importance(xg_reg)
 #matplotlib.pyplot.rcParams['figure.figsize'] = [5, 5]
 #matplotlib.pyplot.show()
+
+#xg_reg = xgboost.train(params=params, dtrain=data_dmatrix, num_boost_round=10, callbacks=[wandb_callback()])
+
+# MLOPS - W&B analytics
+# added the wandb to the callbacks