Remove add_eot and add_sot parameters from training configuration

dan/manager/ocr.py

@@ -174,11 +174,9 @@ class OCRDataset(GenericDataset):
 
         sample["label"] = full_label
         sample["token_label"] = token_to_ind(self.charset, full_label)
-        if "add_eot" in self.params["config"]["constraints"]:
-            sample["token_label"].append(self.tokens["end"])
+        sample["token_label"].append(self.tokens["end"])
         sample["label_len"] = len(sample["token_label"])
-        if "add_sot" in self.params["config"]["constraints"]:
-            sample["token_label"].insert(0, self.tokens["start"])
+        sample["token_label"].insert(0, self.tokens["start"])
         return sample
 
 

dan/ocr/document/train.py

@@ -109,10 +109,7 @@ def get_config():
                 "height_divisor": 32,  # Image height will be divided by 32
                 "padding_value": 0,  # Image padding value
                 "padding_token": None,  # Label padding value
-                "constraints": [
-                    "add_eot",
-                    "add_sot",
-                ],  # add end-of-transcription and start-of-transcription tokens in labels
+                "constraints": [],
                 "preprocessings": [
                     {
                         "type": "to_RGB",

tests/conftest.py

@@ -72,10 +72,7 @@ def training_config():
                 "height_divisor": 32,  # Image height will be divided by 32
                 "padding_value": 0,  # Image padding value
                 "padding_token": None,  # Label padding value
-                "constraints": [
-                    "add_eot",
-                    "add_sot",
-                ],  # add end-of-transcription and start-of-transcription tokens in labels
+                "constraints": [],
                 "preprocessings": [
                     {
                         "type": "to_RGB",