Fix text classification data loading

CHANGELOG.md

@@ -41,6 +41,8 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 
 - Fixed a bug where additional kwargs (e.g. sampler) passed to tabular data would be ignored ([#792](https://github.com/PyTorchLightning/lightning-flash/pull/792))
 
+- Fixed a bug where loading text data with additional non-numeric columns (not input or target) would give an error ([#888](https://github.com/PyTorchLightning/lightning-flash/pull/888))
+
 
 ## [0.5.0] - 2021-09-07
 

flash/text/classification/data.py

@@ -132,11 +132,14 @@ def load_data(
                 # rename label column
                 hf_dataset = hf_dataset.rename_column(target, DefaultDataKeys.TARGET)
 
+        # remove extra columns
+        extra_columns = set(hf_dataset.column_names) - {input, DefaultDataKeys.TARGET}
+        hf_dataset = hf_dataset.remove_columns(extra_columns)
+
         # tokenize
-        hf_dataset = hf_dataset.map(partial(self._tokenize_fn, input=input), batched=True)
+        hf_dataset = hf_dataset.map(partial(self._tokenize_fn, input=input), batched=True, remove_columns=[input])
 
         # set format
-        hf_dataset = hf_dataset.remove_columns([input])  # just leave the numerical columns
         hf_dataset.set_format("torch")
 
         return hf_dataset

tests/text/classification/test_model.py

@@ -95,8 +95,7 @@ def test_load_from_checkpoint_dependency_error():
     "cli_args",
     (
         ["flash", "text_classification", "--trainer.fast_dev_run", "True"],
-        # TODO: update this to work with Pietro's new text data loading (separate PR)
-        # ["flash", "text_classification", "--trainer.fast_dev_run", "True", "from_toxic"],
+        ["flash", "text_classification", "--trainer.fast_dev_run", "True", "from_toxic"],
     ),
 )
 def test_cli(cli_args):