Apply 0c2612fd

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",

docs/usage/train/parameters.md

@@ -8,7 +8,7 @@ All hyperparameters are specified and editable in the training scripts (meaning
 | `dataset_name`                          | Name of the dataset.                                                                   | `str`        |                                                |
 | `dataset_level`                         | Level of the dataset. Should be named after the element type.                          | `str`        |                                                |
 | `dataset_variant`                       | Variant of the dataset. Usually empty for HTR datasets, `"_sem"` for HTR+NER datasets. | `str`        |                                                |
-| `dataset_path`                          | Path to the dataset.                                                                   | `str`        |
+| `dataset_path`                          | Path to the dataset.                                                                   | `str`        |                                                |
 | `dataset_params.config.dataset_manager` | Dataset manager class.                                                                 | custom class | `OCRDatasetManager`                            |
 | `dataset_params.config.dataset_class`   | Dataset class.                                                                         | custom class | `OCRDataset`                                   |
 | `dataset_params.config.datasets`        | Dataset dictionary with the dataset name as key and dataset path as value.             | `dict`       |                                                |
@@ -18,7 +18,7 @@ All hyperparameters are specified and editable in the training scripts (meaning
 | `dataset_params.config.width_divisor`   | Factor to reduce the height of the feature vector before feeding the decoder.          | `int`        | `32`                                           |
 | `dataset_params.config.padding_value`   | Image padding value.                                                                   | `int`        | `0`                                            |
 | `dataset_params.config.padding_token`   | Transcription padding value.                                                           | `int`        | `None`                                         |
-| `dataset_params.config.constraints`     | Whether to add end-of-transcription and start-of-transcription tokens in labels.       | `list`       | `["add_eot", "add_sot"]`                       |
+| `dataset_params.config.constraints`     | Whether to add end-of-transcription and start-of-transcription tokens in labels.       | `list`       | `[]`                                           |
 | `dataset_params.config.preprocessings`  | List of pre-processing functions to apply to input images.                             | `list`       | (see [dedicated section](#data-preprocessing)) |
 | `dataset_params.config.augmentation`    | Configuration for data augmentation.                                                   | `dict`       | (see [dedicated section](#data-augmentation))  |
 

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",