Add more behaviour (continuous training)

spleenseg/spleenseg.py

@@ -17,6 +17,7 @@
 from typing import Any, Optional, Callable
 
 from spleenseg.core import neuralnet
+from spleenseg.models.data import TrainingParams
 from spleenseg.transforms import transforms
 from spleenseg.plotting import plotting
 import warnings
@@ -56,6 +57,12 @@
     default="training",
     help="mode of behaviour: training or inference",
 )
+parser.add_argument(
+    "--logTrainingTransformVols",
+    default=False,
+    action="store_true",
+    help="If specified, save training tranformations as NIfTI volumes",
+)
 parser.add_argument(
     "--useModel",
     type=str,
@@ -171,12 +178,21 @@ def envDetail_print(options: Namespace, **kwargs):
 
 def env_outputDirsMake(options: Namespace) -> None:
     if "training" in options.mode:
-        Path(Path(options.outputdir) / "training").mkdir(parents=True, exist_ok=True)
-        Path(Path(options.outputdir) / "validation").mkdir(parents=True, exist_ok=True)
+        params: TrainingParams = TrainingParams(options)
+        params.whileTrainingIO.mkdir(parents=True, exist_ok=True)
+        params.whileTrainingValidation.mkdir(parents=True, exist_ok=True)
+        params.postTrainingValidation.mkdir(parents=True, exist_ok=True)
     if "inference" in options.mode:
         Path(Path(options.outputdir) / "inference").mkdir(parents=True, exist_ok=True)
 
 
+def modelFile_inputdirGet(options: Namespace) -> Path:
+    modelFile: Path = Path(Path(options.inputdir) / options.useModel)
+    if not modelFile.exists():
+        raise FileNotFoundError(f"The model '{modelFile}' does not exist.")
+    return modelFile
+
+
 def training_do(neuralNet: neuralnet.NeuralNet, options: Namespace) -> bool:
     trainingOK: bool = True
 
@@ -189,6 +205,8 @@ def training_do(neuralNet: neuralnet.NeuralNet, options: Namespace) -> bool:
 
     if options.mode == "training":
         neuralNet.train()
+    if options.mode == "trainingContinue":
+        neuralNet.train(modelFile_inputdirGet(options))
 
     plotting.plot_trainingMetrics(
         neuralNet.trainingLog,
@@ -206,11 +224,7 @@ def inference_do(neuralNet: neuralnet.NeuralNet, options: Namespace) -> bool:
     inferenceOK: bool = True
     neuralNet.testingFileSet = testingData_prep(options)
 
-    modelFile: Path = Path(Path(options.inputdir) / options.useModel)
-    if not modelFile.exists():
-        raise FileNotFoundError(f"The model '{modelFile}' does not exist.")
-
-    neuralNet.infer_usingModel(modelFile)
+    neuralNet.infer_usingModel(modelFile_inputdirGet(options))
 
     return inferenceOK