dont use .convert_dataset in predict script (#214)

ArneBinder · web-flow · commit f0f4dcfa51c4 · 2025-09-24T18:45:17.000+02:00
diff --git a/src/predict.py b/src/predict.py
@@ -94,7 +94,7 @@ def predict(cfg: DictConfig) -> Tuple[dict, dict]:
             )
 
         # auto-convert the dataset if the taskmodule specifies a document type
-        dataset = pipeline.taskmodule.convert_dataset(dataset)
+        dataset = dataset.to_document_type(pipeline.taskmodule, downcast=False)
 
     # Init the serializer
     serializer: Optional[DocumentSerializer] = None

Original file line number	Diff line number	Diff line change
`@@ -94,7 +94,7 @@ def predict(cfg: DictConfig) -> Tuple[dict, dict]:`
`94`	`94`	`)`
`95`	`95`
`96`	`96`	`# auto-convert the dataset if the taskmodule specifies a document type`
`97`		`- dataset = pipeline.taskmodule.convert_dataset(dataset)`
	`97`	`+ dataset = dataset.to_document_type(pipeline.taskmodule, downcast=False)`
`98`	`98`
`99`	`99`	`# Init the serializer`
`100`	`100`	`serializer: Optional[DocumentSerializer] = None`