Merge branch 'main' into season-glocal-reg

docs/source/tutorials/tutorial10.ipynb

@@ -73,7 +73,7 @@
     "collapsed": false
    },
    "source": [
-    "Validation is performed by passing the validation set to the fit method during training. The resulting metrics show the performance of the model compared to our validation set."
+    "Validation is performed by passing the validation set to the fit method during training. The resulting metrics show the performance of the model compared to our validation set. "
    ]
   },
   {
@@ -356,6 +356,34 @@
     "set_random_seed(0)"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Aditionally, it is important to make sure to set the flag `deterministic` in the `fit` function to `True`."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import pandas as pd\n",
+    "from neuralprophet import NeuralProphet\n",
+    "\n",
+    "# Load the dataset from the CSV file using pandas\n",
+    "df = pd.read_csv(\"https://github.com/ourownstory/neuralprophet-data/raw/main/kaggle-energy/datasets/tutorial01.csv\")\n",
+    "\n",
+    "# Model and prediction\n",
+    "m = NeuralProphet()\n",
+    "\n",
+    "df_train, df_val = m.split_df(df, valid_p=0.2)\n",
+    "\n",
+    "# Set the deterministic flag to True\n",
+    "metrics = m.fit(df_train, validation_df=df_val, progress=None, deterministic=True)"
+   ]
+  },
   {
    "attachments": {},
    "cell_type": "markdown",

neuralprophet/forecaster.py

@@ -905,6 +905,7 @@ def fit(
         checkpointing: bool = False,
         continue_training: bool = False,
         num_workers: int = 0,
+        deterministic: bool = False,
     ):
         """Train, and potentially evaluate model.
 
@@ -1069,6 +1070,7 @@ def fit(
                 checkpointing_enabled=checkpointing,
                 continue_training=continue_training,
                 num_workers=num_workers,
+                deterministic=deterministic,
             )
         else:
             df_val, _, _, _ = df_utils.prep_or_copy_df(validation_df)
@@ -1093,6 +1095,7 @@ def fit(
                 checkpointing_enabled=checkpointing,
                 continue_training=continue_training,
                 num_workers=num_workers,
+                deterministic=deterministic,
             )
 
         # Show training plot
@@ -2714,6 +2717,7 @@ def _train(
         checkpointing_enabled: bool = False,
         continue_training=False,
         num_workers=0,
+        deterministic: bool = False,
     ):
         """
         Execute model training procedure for a configured number of epochs.
@@ -2771,6 +2775,7 @@ def _train(
             metrics_enabled=metrics_enabled,
             checkpointing_enabled=checkpointing_enabled,
             num_batches_per_epoch=len(train_loader),
+            deterministic=deterministic,
         )
 
         # Tune hyperparams and train

neuralprophet/utils.py

@@ -11,6 +11,7 @@
 import pandas as pd
 import pytorch_lightning as pl
 import torch
+from lightning_fabric.utilities.seed import seed_everything
 
 from neuralprophet import utils_torch
 from neuralprophet.logger import ProgressBar
@@ -710,6 +711,7 @@ def set_random_seed(seed: int = 0):
     """
     np.random.seed(seed)
     torch.manual_seed(seed)
+    seed_everything(seed, workers=True)
 
 
 def set_logger_level(logger, log_level, include_handlers=False):
@@ -818,6 +820,7 @@ def configure_trainer(
     metrics_enabled: bool = False,
     checkpointing_enabled: bool = False,
     num_batches_per_epoch: int = 100,
+    deterministic: bool = False,
 ):
     """
     Configures the PyTorch Lightning trainer.
@@ -888,6 +891,8 @@ def configure_trainer(
     else:
         config["logger"] = False
 
+    config["deterministic"] = deterministic
+
     # Configure callbacks
     callbacks = []
     has_custom_callbacks = True if "callbacks" in config else False