BART for summarization training with CNN/DM using pytorch-lightning

examples/transformer_base.py

@@ -53,10 +53,9 @@ class BaseTransformer(pl.LightningModule):
         super(BaseTransformer, self).__init__()
         self.hparams = hparams
         self.hparams.model_type = self.hparams.model_type.lower()
-
         config = AutoConfig.from_pretrained(
             self.hparams.config_name if self.hparams.config_name else self.hparams.model_name_or_path,
-            num_labels=num_labels,
+            **({"num_labels": num_labels} if num_labels is not None else {}),
             cache_dir=self.hparams.cache_dir if self.hparams.cache_dir else None,
         )
         tokenizer = AutoTokenizer.from_pretrained(