update hydronaut experiments with decorator

experiments_hydronaut/experiment.py

 from hydronaut.experiment import Experiment
+from hydronaut.decorator import with_hydronaut
+
 import sys
 sys.path.append(".")
 sys.path.append("competitors/swotted")
@@ -41,7 +43,7 @@ class SWoTTeDExperiment(Experiment):
         fig, axs = plt.subplots(R, 2)
         for i in range(R):
             axs[i, 0].imshow(
-                Ph_[i], cmap="gray", vmin=0, vmax=1, interpolation="nearest"
+                Ph_[i].detach().numpy(), cmap="gray", vmin=0, vmax=1, interpolation="nearest"
             )
             axs[i, 0].set_ylabel("Drugs")
             axs[i, 0].set_xlabel("time")
@@ -68,7 +70,12 @@ class SyntheticDatasetExperiment(SWoTTeDExperiment):
     All the parameters of the experiments (experiments, dataset and SWoTTeD parameters)
     have to be located in the `conf` directory
 
-    To run this experiment,:
+    To run this experiment, you simply have to run the script:
+    ```bash
+    python experiment.py
+    ```
+
+    To run this experiment from the hydronaut commands:
     ```bash
     export HYDRONAUT_CONFIG=experiments_hydronaut/experiment_synthetic.yaml
     hydronaut-run```
@@ -150,111 +157,11 @@ class SyntheticDatasetExperiment(SWoTTeDExperiment):
 
         return ret[0]["test_loss"]
 
-class HyperparametersTuning(SWoTTeDExperiment):
-    """Hydronaut experiment representing an experiment to fine tune a SWoTTeD model 
-    on a datasets.
-    
-    To run this experiment,:
-    ```bash
-    export HYDRONAUT_CONFIG=experiments_hydronaut/experiment_tuning.yaml
-    hydronaut-run```
-
-    """
-
-    def gen_phenotypes_image(self, filename="phenotypes.png"):
-        """Create a PNG image to compare the extracted phenotypes.
-
-        This function reorders the phenotypes.
-
-        Parameters
-        ==========
-        filename: str
-            name of the file to generate (default: 'phenotype.png')"""
-        R=len(self.swotted.Ph)
-        fig, axs = plt.subplots(R,1)
-        for i in range(R):
-            axs[i].imshow(
-                self.swotted.Ph[i].detach().numpy(), cmap="gray", vmin=0, vmax=1, interpolation="nearest"
-            )
-            axs[i].set_ylabel("Features")
-            axs[i].set_xlabel("Time")
-        fig.savefig(filename)
-
-    def __call__(self) -> float:
-        """Implement the detail of one instance of the Hydronaut experiment.
-
-        This experiment generates several artifacts:
-        - `Ph.pkl`: the extracted phenotypes (the model)
-        - ̀ phenotypes.png`: illustration of the phenotypes
-
-        Returns
-        --------
-        float
-            Test loss metric
-        """
-
-        params = self.config.experiment.params
-        
-        _, _, X, _ = pickle.load(open(params.data.filename, "rb"))
-        if isinstance(X,list):
-            nb_train = int(params.data.percent_train*len(X))
-            nb_test = len(X)-nb_train
-
-            #remove examples that do not have the correct length
-            to_keep=[]
-            for i in range(len(X)):
-                if X[i].shape[1]>=params.model.twl:
-                    to_keep.append(i)
-            if len(to_keep)<len(X):
-                warnings.warn(f"Removed {len(X)-len(to_keep)} examples with incompatible length")
-                X=[X[i] for i in to_keep]
-
-        else:
-            nb_train = int(params.data.percent_train*X.shape[0])
-            nb_test = X.shape[0]-nb_train
-
-        with open_dict(params):
-            params.model.N = X[0].shape[0]
-
-        # define the model
-        self.swotted = swottedModule(params)
-
-        train_data_loader = DataLoader(
-            Subset(X[: nb_train], np.arange(nb_train)),
-            batch_size=params.training.batch_size,
-            shuffle=False,
-            collate_fn=lambda x: x,
-        )
-        test_data_loader = DataLoader(
-            Subset(X[nb_train :], np.arange(len(X[nb_train :]))),
-            batch_size=params.training.batch_size,
-            shuffle=False,
-            collate_fn=lambda x: x,
-        )
-
-        trainer = swottedTrainer(max_epochs=params.training.nepochs)
-        before = time.time()
-        trainer.fit(model=self.swotted, train_dataloaders=train_data_loader)
-        duration = time.time() - before
-        self.log_metric("training_time", duration)
-
-        # save model
-        pickle.dump(self.swotted.Ph, open("Ph.pkl", "wb"))
-        self.log_artifact("Ph.pkl", "Ph_model")
-
-        self.gen_phenotypes_image("phenotypes.png")
-        self.log_artifact("phenotypes.png", "Ph_images")
-
-        ret = trainer.test(self.swotted, test_data_loader)
-        # ret is a (list of) dictionary that contains the logged values.
-        
-        # Log the both metrics
-        self.log_metric("test_loss", ret[0]["test_loss"])
-
-        return ret[0]["test_loss"]
+@with_hydronaut(config_path='experiments_hydronaut/experiment_synthetic.yaml')
+def main(config):
+    experiment=SyntheticDatasetExperiment(config)
+    return experiment()
 
-# run Hydronaut with the default parameters
-from hydronaut.run import Runner
 
-if __name__ == "__main__":
-    Runner()()
+if __name__ == '__main__':
+    sys.exit(main())
\ No newline at end of file

experiments_hydronaut/tuning.py

+from hydronaut.experiment import Experiment
+from hydronaut.decorator import with_hydronaut
+
+import sys
+sys.path.append(".")
+sys.path.append("competitors/swotted")
+
+from torch.utils.data import DataLoader
+from swotted.swotted import swottedModule, swottedTrainer
+from swotted.utils import Subset
+from swotted.loss_metrics import *
+from experiments.gen_data import gen_synthetic_data
+from experiments_hydronaut.experiment import SWoTTeDExperiment
+
+from omegaconf import open_dict
+
+import numpy as np
+import time
+import matplotlib.pyplot as plt
+
+import pickle
+
+import warnings
+
+
+class HyperparametersTuning(SWoTTeDExperiment):
+    """Hydronaut experiment representing an experiment to fine tune a SWoTTeD model 
+    on a datasets.
+
+    To run this experiment from hyronaut script:
+    ```bash
+    export HYDRONAUT_CONFIG=experiments_hydronaut/experiment_tuning.yaml
+    hydronaut-run```
+
+    """
+
+    def gen_phenotypes_image(self, filename="phenotypes.png"):
+        """Create a PNG image to compare the extracted phenotypes.
+
+        This function reorders the phenotypes.
+
+        Parameters
+        ==========
+        filename: str
+            name of the file to generate (default: 'phenotype.png')"""
+        R=len(self.swotted.Ph)
+        fig, axs = plt.subplots(R,1)
+        for i in range(R):
+            axs[i].imshow(
+                self.swotted.Ph[i].detach().numpy(), cmap="gray", vmin=0, vmax=1, interpolation="nearest"
+            )
+            axs[i].set_ylabel("Features")
+            axs[i].set_xlabel("Time")
+        fig.savefig(filename)
+
+    def __call__(self) -> float:
+        """Implement the detail of one instance of the Hydronaut experiment.
+
+        This experiment generates several artifacts:
+        - `Ph.pkl`: the extracted phenotypes (the model)
+        - ̀ phenotypes.png`: illustration of the phenotypes
+
+        Returns
+        --------
+        float
+            Test loss metric
+        """
+
+        params = self.config.experiment.params
+        
+        _, _, X, _ = pickle.load(open(params.data.filename, "rb"))
+        if isinstance(X,list):
+            nb_train = int(params.data.percent_train*len(X))
+            nb_test = len(X)-nb_train
+
+            #remove examples that do not have the correct length
+            to_keep=[]
+            for i in range(len(X)):
+                if X[i].shape[1]>=params.model.twl:
+                    to_keep.append(i)
+            if len(to_keep)<len(X):
+                warnings.warn(f"Removed {len(X)-len(to_keep)} examples with incompatible length")
+                X=[X[i] for i in to_keep]
+
+        else:
+            nb_train = int(params.data.percent_train*X.shape[0])
+            nb_test = X.shape[0]-nb_train
+
+        with open_dict(params):
+            params.model.N = X[0].shape[0]
+
+        # define the model
+        self.swotted = swottedModule(params)
+
+        train_data_loader = DataLoader(
+            Subset(X[: nb_train], np.arange(nb_train)),
+            batch_size=params.training.batch_size,
+            shuffle=False,
+            collate_fn=lambda x: x,
+        )
+        test_data_loader = DataLoader(
+            Subset(X[nb_train :], np.arange(len(X[nb_train :]))),
+            batch_size=params.training.batch_size,
+            shuffle=False,
+            collate_fn=lambda x: x,
+        )
+
+        trainer = swottedTrainer(fast_dev_run=False, max_epochs=params.training.nepochs)
+        before = time.time()
+        trainer.fit(model=self.swotted, train_dataloaders=train_data_loader)
+        duration = time.time() - before
+        self.log_metric("training_time", duration)
+
+        # save model
+        pickle.dump(self.swotted.Ph, open("Ph.pkl", "wb"))
+        self.log_artifact("Ph.pkl", "Ph_model")
+
+        self.gen_phenotypes_image("phenotypes.png")
+        self.log_artifact("phenotypes.png", "Ph_images")
+
+        ret = trainer.test(self.swotted, test_data_loader)
+        # ret is a (list of) dictionary that contains the logged values.
+        
+        # Log the both metrics
+        self.log_metric("test_loss", ret[0]["test_loss"])
+
+        return ret[0]["test_loss"]
+
+
+@with_hydronaut(config_path='experiments_hydronaut/experiment_tuning.yaml')
+def main(config):
+    experiment=HyperparametersTuning(config)
+    return experiment()
+
+if __name__ == '__main__':
+    sys.exit(main())