Bug fixes

avae/config.py

@@ -1,5 +1,6 @@
 import logging
 import os
+import pathlib
 
 import yaml
 from pydantic import (
@@ -110,7 +111,7 @@ class AffinityConfig(BaseModel):
     )
     pose_dims: int = Field(1, description="Pose dimensions")
 
-    rescale: float = Field(None, description="Rescale data")
+    rescale: int | None = Field(None, description="Rescale data")
     restart: bool = Field(False, description="Restart training")
     shift_min: bool = Field(
         False, description="Scale data with min-max transformation"
@@ -291,12 +292,14 @@ def load_config_params(
                     + key
                     + " in config file or command line arguments. Default values will be used."
                 )
-
+        # turning all path into absolute paths and strings for output saving
+        if type(val) == pathlib.PosixPath:
+            setattr(data, key, str(val.absolute()))
     # return data as dictionary
     return data.model_dump()
 
 
-def write_config_file(time_stamp_name, data):
+def write_config_file(time_stamp_name, data) -> None:
     # record final configuration in logger and save to yaml file
     for key, val in data.items():
         logging.info("Parameter " + key + " set to value: " + str(data[key]))
@@ -310,31 +313,62 @@ def write_config_file(time_stamp_name, data):
     logging.info("YAML File saved!\n")
 
 
-def setup_visualisation_config(data: dict) -> None:
-    settings.VIS_LOS = (
+def setup_visualisation_config(data: dict) -> dict:
+    """
+    Set up visualisation configurations. Logic is the following: if a specific visualisation is not set, it will be set to the value of vis_all.
+    For frequency, if it is set to 0 (the default), it will be set to the value of freq_all.
+
+    Parameters
+    ----------
+    data : dict
+        Dictionary of configuration parameters.
+
+    Returns
+    -------
+    data : dict
+        Dictionary of configuration parameters updated with logic.
+    """
+    data["vis_los"] = (
         data["vis_los"] if data["vis_los"] is not None else data["vis_all"]
     )
-    settings.VIS_ACC = (
+    settings.VIS_LOS = data["vis_los"]
+
+    data["vis_acc"] = (
         data["vis_acc"] if data["vis_acc"] is not None else data["vis_all"]
     )
-    settings.VIS_REC = (
+
+    settings.VIS_ACC = data["vis_acc"]
+
+    data["vis_rec"] = (
         data["vis_rec"] if data["vis_rec"] is not None else data["vis_all"]
     )
-    settings.VIS_CYC = (
+    settings.VIS_REC = data["vis_rec"]
+
+    data["vis_cyc"] = (
         data["vis_cyc"] if data["vis_cyc"] is not None else data["vis_all"]
     )
-    settings.VIS_AFF = (
+    settings.VIS_CYC = data["vis_cyc"]
+
+    data["vis_aff"] = (
         data["vis_aff"] if data["vis_aff"] is not None else data["vis_all"]
     )
-    settings.VIS_EMB = (
+    settings.VIS_AFF = data['vis_aff']
+
+    data["vis_emb"] = (
         data["vis_emb"] if data["vis_emb"] is not None else data["vis_all"]
     )
-    settings.VIS_INT = (
+    settings.VIS_EMB = data["vis_emb"]
+
+    data["vis_int"] = (
         data["vis_int"] if data["vis_int"] is not None else data["vis_all"]
     )
-    settings.VIS_DIS = (
+    settings.VIS_INT = data["vis_int"]
+
+    data["vis_dis"] = (
         data["vis_dis"] if data["vis_dis"] is not None else data["vis_all"]
     )
+    settings.VIS_DIS = data["vis_dis"]
+
     settings.VIS_POS = (
         data["vis_pos"] if data["vis_pos"] is not None else data["vis_all"]
     )
@@ -351,30 +385,49 @@ def setup_visualisation_config(data: dict) -> None:
     settings.VIS_FORMAT = data["vis_format"]
     settings.VIS_Z_N_INT = data["vis_z_n_int"]
 
-    settings.FREQ_EVAL = (
+    data["freq_eval"] = (
         data["freq_eval"] if data["freq_eval"] != 0 else data["freq_all"]
     )
-    settings.FREQ_REC = (
+    settings.FREQ_EVAL = data["freq_eval"]
+
+    data["freq_rec"] = (
         data["freq_rec"] if data["freq_rec"] != 0 else data["freq_all"]
     )
-    settings.FREQ_EMB = (
+    settings.FREQ_REC = data["freq_rec"]
+
+    data["freq_emb"] = (
         data["freq_emb"] if data["freq_emb"] != 0 else data["freq_all"]
     )
-    settings.FREQ_INT = (
+    settings.FREQ_EMB = data["freq_emb"]
+
+    data["freq_int"] = (
         data["freq_int"] if data["freq_int"] != 0 else data["freq_all"]
     )
-    settings.FREQ_DIS = (
+    settings.FREQ_INT = data["freq_int"]
+
+    data["freq_dis"] = (
         data["freq_dis"] if data["freq_dis"] != 0 else data["freq_all"]
     )
-    settings.FREQ_POS = (
+    settings.FREQ_DIS = data["freq_dis"]
+
+    data["freq_pos"] = (
         data["freq_pos"] if data["freq_pos"] != 0 else data["freq_all"]
     )
-    settings.FREQ_ACC = (
+    settings.FREQ_POS = data["freq_pos"]
+
+    data["freq_acc"] = (
         data["freq_acc"] if data["freq_acc"] != 0 else data["freq_all"]
     )
-    settings.FREQ_STA = (
+    settings.FREQ_ACC = data["freq_acc"]
+
+    data['freq_sta'] = (
         data["freq_sta"] if data["freq_sta"] != 0 else data["freq_all"]
     )
-    settings.FREQ_SIM = (
+    settings.FREQ_STA = data["freq_sta"]
+
+    data["freq_sim"] = (
         data["freq_sim"] if data["freq_sim"] != 0 else data["freq_all"]
     )
+    settings.FREQ_SIM = data["freq_sim"]
+
+    return data

avae/data.py

@@ -33,7 +33,7 @@ def load_data(
     gaussian_blur: bool = False,
     normalise: bool = False,
     shift_min: bool = False,
-    rescale: bool | None = None,
+    rescale: int | None = None,
 ) -> tuple[DataLoader, int]:
     ...
 
@@ -53,7 +53,7 @@ def load_data(
     gaussian_blur: bool = False,
     normalise: bool = False,
     shift_min: bool = False,
-    rescale: bool | None = None,
+    rescale: int | None = None,
 ) -> tuple[DataLoader, DataLoader, DataLoader, pd.DataFrame, int]:
     ...
 
@@ -72,7 +72,7 @@ def load_data(
     gaussian_blur: bool = False,
     normalise: bool = False,
     shift_min: bool = False,
-    rescale: bool | None = None,
+    rescale: int | None = None,
 ) -> tuple[DataLoader, DataLoader, DataLoader, pd.DataFrame, int] | tuple[
     DataLoader, int
 ]:
@@ -106,7 +106,7 @@ def load_data(
         In True, the input data is normalised before being passed to the model.
     shift_min: bool
         If True, the minimum value of the input data is shifted to 0 and maximum to 1.
-    rescale: int
+    rescale: int | None
         If not None, the input data is rescaled to the given value.
 
 

avae/train.py

@@ -58,7 +58,7 @@ def train(
     gaussian_blur: bool,
     normalise: bool,
     shift_min: bool,
-    rescale: bool,
+    rescale: int | None,
     tensorboard: bool,
     classifier: str,
     strategy: str,
@@ -160,8 +160,8 @@ def train(
         Path to the beta values to load.
     gamma_load: str
         Path to the gamma values to load.
-    rescale: bool
-        If True, the input data is rescaled to have a mean of 0 and std of 1.
+    rescale: int | None
+        If provided, the data is rescaled by the value.
     """
     lt.pytorch.seed_everything(42)
 

run.py

@@ -661,7 +661,7 @@ def run(
         logging.getLogger("matplotlib.font_manager").disabled = True
 
     # visualisation global settings defined from config file
-    setup_visualisation_config(data)
+    data = setup_visualisation_config(data)
 
     if data["new_out"]:
         dir_name = f'results_{settings.date_time_run}_model_{data["model"]}_lat{data["latent_dims"]}_pose{data["pose_dims"]}_lr{data["learning"]}_beta{data["beta"]}_gamma{data["gamma"]}'