nkululeko 0.93.8__tar.gz → 0.93.10__tar.gz

{nkululeko-0.93.8 → nkululeko-0.93.10}/CHANGELOG.md

@@ -1,6 +1,15 @@
 Changelog
 =========
 
+Version 0.93.10
+--------------
+* added nan check for imported features
+* added LOGO result output
+
+Version 0.93.9
+--------------
+* added manual seed to torch models
+
 Version 0.93.8
 --------------
 * fixed bugs in plot

{nkululeko-0.93.8/nkululeko.egg-info → nkululeko-0.93.10}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: nkululeko
-Version: 0.93.8
+Version: 0.93.10
 Summary: Machine learning audio prediction experiments based on templates
 Home-page: https://github.com/felixbur/nkululeko
 Author: Felix Burkhardt
@@ -353,6 +353,15 @@ If you use it, please mention the Nkululeko paper:
 Changelog
 =========
 
+Version 0.93.10
+--------------
+* added nan check for imported features
+* added LOGO result output
+
+Version 0.93.9
+--------------
+* added manual seed to torch models
+
 Version 0.93.8
 --------------
 * fixed bugs in plot

{nkululeko-0.93.8 → nkululeko-0.93.10}/nkululeko/constants.py

@@ -1,2 +1,2 @@
-VERSION="0.93.8"
+VERSION="0.93.10"
 SAMPLING_RATE = 16000

{nkululeko-0.93.8 → nkululeko-0.93.10}/nkululeko/feat_extract/feats_import.py

@@ -30,7 +30,7 @@ class ImportSet(Featureset):
                 "feature type == import needs import_file = ['file1', 'filex']"
             )
         except SyntaxError:
-            if type(feat_import_files) == str:
+            if type(feat_import_files) is str:
                 feat_import_files = [feat_import_files]
             else:
                 self.util.error(f"import_file is wrong: {feat_import_files}")
@@ -40,6 +40,11 @@ class ImportSet(Featureset):
             if not os.path.isfile(feat_import_file):
                 self.util.error(f"no import file: {feat_import_file}")
             df = audformat.utils.read_csv(feat_import_file)
+            if df.isnull().values.any():
+                self.util.warn(
+                    f"imported features contain {df.isna().sum()} NAN, filling with zero."
+                )
+                df = df.fillna(0)
             df = self.util.make_segmented_index(df)
             df = df[df.index.isin(self.data_df.index)]
             if import_files_append:

{nkululeko-0.93.8 → nkululeko-0.93.10}/nkululeko/models/model.py

@@ -171,6 +171,7 @@ class Model:
             f"LOGO: {self.logo} folds: mean {results.mean():.3f}, std:"
             f" {results.std():.3f}"
         )
+        report.print_logo(results)
 
     def train(self):
         """Train the model."""

{nkululeko-0.93.8 → nkululeko-0.93.10}/nkululeko/models/model_mlp.py

@@ -31,6 +31,10 @@ class MLPModel(Model):
         super().set_model_type("ann")
         self.name = "mlp"
         self.target = glob_conf.config["DATA"]["target"]
+        manual_seed = eval(self.util.config_val("MODEL", "manual_seed", "True"))
+        if manual_seed:
+            self.util.debug(f"seeding random to {23}")
+            torch.manual_seed(23)
         labels = glob_conf.labels
         self.class_num = len(labels)
         # set up loss criterion

{nkululeko-0.93.8 → nkululeko-0.93.10}/nkululeko/models/model_mlp_regression.py

@@ -36,6 +36,10 @@ class MLP_Reg_model(Model):
         else:
             self.util.error(f"unknown loss function: {criterion}")
         self.util.debug(f"training model with {criterion} loss function")
+        manual_seed = eval(self.util.config_val("MODEL", "manual_seed", "True"))
+        if manual_seed:
+            self.util.debug(f"seeding random to {23}")
+            torch.manual_seed(23)
         # set up the model
         cuda = "cuda" if torch.cuda.is_available() else "cpu"
         self.device = self.util.config_val("MODEL", "device", cuda)

{nkululeko-0.93.8 → nkululeko-0.93.10}/nkululeko/reporting/reporter.py

@@ -380,6 +380,18 @@ class Reporter:
     def set_filename_add(self, my_string):
         self.filenameadd = f"_{my_string}"
 
+    def print_logo(self, results):
+        res_dir = self.util.get_path("res_dir")
+        result_str = f"LOGO results: [{','.join(results.astype(str))}]"
+        file_name = f"{res_dir}/logo_results.txt"
+        with open(file_name, "w") as text_file:
+            text_file.write(
+                f"LOGO: mean {results.mean():.3f},  std: " + f"{results.std():.3f}"
+            )
+            text_file.write("\n")
+            text_file.write(result_str)
+        self.util.debug(result_str)
+
     def print_results(self, epoch=None):
         if epoch is None:
             epoch = self.epoch

{nkululeko-0.93.8 → nkululeko-0.93.10/nkululeko.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: nkululeko
-Version: 0.93.8
+Version: 0.93.10
 Summary: Machine learning audio prediction experiments based on templates
 Home-page: https://github.com/felixbur/nkululeko
 Author: Felix Burkhardt
@@ -353,6 +353,15 @@ If you use it, please mention the Nkululeko paper:
 Changelog
 =========
 
+Version 0.93.10
+--------------
+* added nan check for imported features
+* added LOGO result output
+
+Version 0.93.9
+--------------
+* added manual seed to torch models
+
 Version 0.93.8
 --------------
 * fixed bugs in plot