Deprecate --existing_model, make model library more lenient

configs/default.json

@@ -56,7 +56,6 @@
             "normalization_file": null
         },
         "model": {
-            "existing_model": null,
             "freeze_conv_layers": false,
             "freeze_dense_layers": false,
             "freeze_recurrent_layers": false,

configs/finetuning.json

@@ -36,10 +36,10 @@
             "normalization_file": "/path/to/normalization.json"
         },
         "model": {
-            "existing_model": "/path/to/model/",
             "freeze_conv_layers": true,
             "freeze_dense_layers": false,
             "freeze_recurrent_layers": false,
+            "model": "/path/to/model/",
             "model_name": "My-Finetuned-Loghi-Model",
             "replace_final_layer": true,
             "replace_recurrent_layer": null,

configs/inference.json

@@ -21,7 +21,7 @@
             "deterministic": false
         },
         "model": {
-            "existing_model": "/path/to/model/",
+            "model": "/path/to/model/",
             "use_float32": false
         }
     }

configs/testing.json

@@ -15,7 +15,7 @@
             "normalization_file": "/path/to/normalization.json"
         },
         "model": {
-            "existing_model": "/path/to/model/",
+            "model": "/path/to/model/",
             "use_float32": false
         },
         "training": {

configs/validation.json

@@ -15,7 +15,7 @@
             "normalization_file": "/path/to/normalization.json"
         },
         "model": {
-            "existing_model": "/path/to/model/",
+            "model": "/path/to/model/",
             "use_float32": false
         },
         "training": {

src/main.py

@@ -49,9 +49,9 @@ def main():
         os.makedirs(config["output"], exist_ok=True)
 
     # Get the initial character list
-    if config["existing_model"] or config["charlist"]:
+    if os.path.isdir(config["model"]) or config["charlist"]:
         charlist, removed_padding = load_initial_charlist(
-            config["charlist"], config["existing_model"],
+            config["charlist"], config["model"],
             config["output"], config["replace_final_layer"])
     else:
         charlist = []

src/model/management.py

@@ -86,7 +86,7 @@ def customize_model(model: tf.keras.Model,
                                         use_mask=config["use_mask"])
 
     # Replace the final layer if specified
-    if config["replace_final_layer"] or not config["existing_model"]:
+    if config["replace_final_layer"] or not os.path.isdir(config["model"]):
         new_classes = len(charlist) + \
             2 if config["use_mask"] else len(charlist) + 1
         logging.info("Replacing final layer with %s classes", new_classes)
@@ -191,9 +191,10 @@ def load_or_create_model(config: Config,
         The loaded or newly created Keras model.
     """
 
-    if config["existing_model"]:
-        model = load_model_from_directory(
-            config["existing_model"], custom_objects=custom_objects)
+    # Check if config["model"] is a directory
+    if os.path.isdir(config["model"]):
+        model = load_model_from_directory(config["model"],
+                                          custom_objects=custom_objects)
         if config["model_name"]:
             model._name = config["model_name"]
     else:

src/model/vgsl_model_generator.py

@@ -108,7 +108,7 @@ def __init__(self,
             raise ValueError("No model provided. Please provide a model name "
                              "from the model library or a VGSL-spec string.")
 
-        if model_spec.startswith("model"):
+        if model_spec in self.model_library.keys():
             try:
                 logging.info("Pulling model from model library")
                 model_string = self.model_library[model_spec]

src/setup/arg_parser.py

@@ -137,10 +137,6 @@ def get_arg_parser():
     model_args.add_argument('--use_float32', action='store_true',
                             help="Use 32-bit float precision in the model. "
                             "Can improve performance at the cost of memory.")
-    model_args.add_argument('--existing_model', metavar='existing_model',
-                            type=str, default=None, help="Path to an existing "
-                            "model to continue training, validation, testing, "
-                            "or inferencing. Used as a starting point.")
     model_args.add_argument('--model_name', metavar='model_name', type=str,
                             default=None, help="Custom name for the model, "
                             "used in outputs. Default: None (uses the model "
@@ -273,20 +269,26 @@ def get_arg_parser():
     depr_args.add_argument('--thaw', action='store_true',
                            help="Unfreeze convolutional layers in an "
                            "existing model for further training.")
+    depr_args.add_argument('--existing_model', metavar='existing_model',
+                           type=str, default=None, help="Path to an existing "
+                           "model to continue training, validation, testing, "
+                           "or inferencing. Used as a starting point.")
 
     return parser
 
 
 def fix_args(args):
     if not args.no_auto and args.train_list:
-        logging.warning('do_train implied by providing a train_list')
+        logging.warning('--do_train implied by providing a train_list')
         args.__dict__['do_train'] = True
     if not args.no_auto and args.batch_size > 1:
-        logging.warning('batch_size > 1, setting use_mask=True')
+        logging.warning('--batch_size > 1, setting use_mask=True')
         args.__dict__['use_mask'] = True
     if not args.no_auto and args.inference_list:
-        logging.warning('do_inference implied by providing a inference_list')
+        logging.warning('--do_inference implied by providing a inference_list')
         args.__dict__['do_inference'] = True
+    if not args.no_auto and args.existing_model:
+        args.__dict__['model'] = args.existing_model
 
 
 def arg_future_warning(args):
@@ -322,6 +324,10 @@ def arg_future_warning(args):
     if args.thaw:
         logging.warning("Argument will lose support in May 2024: --thaw. "
                         "Models are saved with all layers thawed by default.")
+    if args.existing_model:
+        logger.warning("Argument will lose support in May 2024: "
+                       "--existing_model. The --model argument can be used "
+                       "to load or create a model instead.")
 
 
 def get_args():

src/setup/config.py

@@ -191,7 +191,6 @@ def organize_args(self, args: argparse.Namespace) -> dict:
             "model": {
                 "model": args.model,
                 "use_float32": args.use_float32,
-                "existing_model": args.existing_model,
                 "model_name": args.model_name,
                 "replace_final_layer": args.replace_final_layer,
                 "replace_recurrent_layer": args.replace_recurrent_layer,
@@ -232,6 +231,7 @@ def organize_args(self, args: argparse.Namespace) -> dict:
                 "output_charlist": args.output_charlist,
                 "config_file_output": args.config_file_output,
                 "thaw": args.thaw,
+                "existing_model": args.existing_model,
             }
         }