Minor fixes

README.md

@@ -22,22 +22,25 @@ See [https://dcase-models.readthedocs.io](https://dcase-models.readthedocs.io/en
 ## Installation instructions
 We recommend to install DCASE-models in a dedicated virtual environment. For instance, using [anaconda](https://www.anaconda.com/):
 ```
-conda create -n dcase python=3.6
+conda create -n dcase python=3.10
 conda activate dcase
 ```
 For GPU support:
 ```
-conda install cudatoolkit cudnn
+conda install cudatoolkit=11.3 cudnn
+conda install cupti=11
+conda install -c nvidia cuda-nvcc
+conda env config vars set LD_LIBRARY_PATH=$CONDA_PREFIX/lib
 ```
 DCASE-models uses [SoX](http://sox.sourceforge.net/) for functions related to the datasets. You can install it in your conda environment by:
 ```
 conda install -c conda-forge sox
 ```
-When installing the library, you must select the tensorflow variant: version 1 (CPU-only or GPU) or version 2.
+When installing the library, you must select the tensorflow variant: version 1 (CPU-only or GPU) or version 2. Note that you can indicate a specific release of DCASE-models (it is possible that you need to pin installation to one of the latest releases if pypi version is not updated)
 ``` 
-pip install DCASE-models[keras_tf] # for tensorflow 1 CPU-only version
-pip install DCASE-models[keras_tf_gpu] # for tensorflow 1 GPU version
-pip install DCASE-models[tf2] # for tensorflow 2
+pip install DCASE-models[keras_tf]==v0.2.0-rc2 # for tensorflow 1 CPU-only version
+pip install DCASE-models[keras_tf_gpu]==v0.2.0-rc2 # for tensorflow 1 GPU version
+pip install DCASE-models[tf2]==v0.2.0-rc2 # for tensorflow 2
 ```
 
 To include visualization related dependencies, run the following instead:

dcase_models/data/feature_extractor.py

@@ -209,9 +209,7 @@ def extract(self, dataset):
                     path_to_features_file = path_audio.replace(
                         audio_path, features_path
                     )
-                    path_to_features_file = path_to_features_file.replace(
-                        'wav', 'npy'
-                    )
+                    path_to_features_file = os.path.splitext(path_to_features_file)[0] + '.npy'
                     np.save(path_to_features_file, features_array)
 
                 # Save parameters.json for future checking
@@ -345,7 +343,7 @@ def pad_audio(self, audio):
             sequence_hop_samples = self.sequence_hop*self.audio_hop
             if len(audio) < sequence_samples:
                 audio = librosa.util.fix_length(
-                    audio, sequence_samples, axis=0, mode=self.pad_mode)
+                    audio, size=sequence_samples, axis=0, mode=self.pad_mode)
             else:
                 if self.sequence_hop_time > 0:
                     audio_frames = int((len(audio) - self.audio_win) / self.audio_hop) + int(((len(audio) - self.audio_win) % self.audio_hop)>0)
@@ -354,7 +352,7 @@ def pad_audio(self, audio):
                     new_samples = new_frames * self.audio_hop + self.audio_win
                     audio = librosa.util.fix_length(
                         audio,
-                        new_samples,
+                        size=new_samples,
                         axis=0, mode=self.pad_mode
                     )
                 else:
@@ -367,8 +365,8 @@ def convert_to_sequences(self, audio_representation):
             audio_representation = np.ascontiguousarray(audio_representation)
             audio_representation = librosa.util.frame(
                 audio_representation,
-                self.sequence_frames,
-                self.sequence_hop,
+                frame_length=self.sequence_frames,
+                hop_length=self.sequence_hop,
                 axis=0
             )
         else:

dcase_models/util/metrics.py

@@ -87,7 +87,7 @@ def evaluate_metrics(model, data, metrics, **kwargs):
         n_files = len(X_val)
         for i in range(n_files):
             X = X_val[i]
-            Y_predicted = model.predict(X)
+            Y_predicted = model.predict(X, verbose=False)
             # if multiple outputs, select the first
             if type(Y_predicted) == list:
                 Y_predicted = Y_predicted[0]
@@ -102,7 +102,7 @@ def evaluate_metrics(model, data, metrics, **kwargs):
             n_files = len(X_val)
             for i in range(n_files):
                 X = X_val[i]
-                Y_predicted = model.predict(X)
+                Y_predicted = model.predict(X, verbose=False)
                 if type(Y_predicted) == list:
                     Y_predicted = Y_predicted[0]
                 predictions.append(Y_predicted)