refactor(attention): removed duplicate exception

BlueCrescent · rrutmann · web-flow · commit f173cffc2da5 · 2026-02-28T14:38:31.000+01:00
Co-authored-by: Richard Rutmann &lt;97447451+rrutmann@users.noreply.github.com&gt;
diff --git a/src/modalities/models/gpt2/gpt2_model.py b/src/modalities/models/gpt2/gpt2_model.py
@@ -532,12 +532,8 @@ def prepare_inter_document_masking(
                 device=device,
             )
             return self._get_unpad_data_for_concatenated_sequences(concatenated_lengths)
-        if self.attention_impl == AttentionImplementation.PYTORCH_FLASH:
-            raise NotImplementedError(
-                "Inter-document masking is not supported for `pytorch_flash`. " "Use `manual` or `dao_flash`."
-            )
         raise NotImplementedError(
-            f"Attention implementation {self.attention_impl} is not supported for inter-document masking."
+            f"Attention implementation {self.attention_impl} is not supported for inter-document masking. Use `manual` or `dao_flash`."
         )
 
     @staticmethod

Original file line number	Diff line number	Diff line change
`@@ -532,12 +532,8 @@ def prepare_inter_document_masking(`
`532`	`532`	`device=device,`
`533`	`533`	`)`
`534`	`534`	`return self._get_unpad_data_for_concatenated_sequences(concatenated_lengths)`
`535`		`- if self.attention_impl == AttentionImplementation.PYTORCH_FLASH:`
`536`		`- raise NotImplementedError(`
`537`		- "Inter-document masking is not supported for `pytorch_flash`. " "Use `manual` or `dao_flash`."
`538`		`- )`
`539`	`535`	`raise NotImplementedError(`
`540`		`- f"Attention implementation {self.attention_impl} is not supported for inter-document masking."`
	`536`	+ f"Attention implementation {self.attention_impl} is not supported for inter-document masking. Use `manual` or `dao_flash`."
`541`	`537`	`)`
`542`	`538`
`543`	`539`	`@staticmethod`