TRF025

Attention masks must be built once in the model, not rebuilt inside a layer or attention module.

Default	Enabled
Scope	Models added on or after 2026-06-20
Source	mlinter/trf025.py
Show in terminal	mlinter --rule TRF025

What it does

In modeling_*.py and modular_*.py, flags calls to a mask factory (the masking_utils entry points create_causal_mask, create_bidirectional_mask, create_sliding_window_causal_mask, create_chunked_causal_mask, create_masks_for_generate, and any create_*_mask helper) inside a class that does not inherit from PreTrainedModel – plain nn.Module blocks such as layers, attention modules and encoders.

Why is this bad?

Mask construction is O(sequence length squared) work that does not vary per layer, so building it in the layer pays that cost once per layer, and each layer then owns its own mask, so the attention backends can no longer be handed a single prepared one. Build it once in the model and pass it down.

Example

 class AcmeLayer(nn.Module):
     def forward(self, hidden_states, attention_mask=None, **kwargs):
-        attention_mask = create_causal_mask(
-            config=self.config, input_embeds=hidden_states, attention_mask=attention_mask, ...
-        )
         return self.self_attn(hidden_states, attention_mask, **kwargs)

 class AcmeModel(AcmePreTrainedModel):
     def forward(self, input_ids=None, attention_mask=None, **kwargs):
+        causal_mask = create_causal_mask(
+            config=self.config, input_embeds=inputs_embeds, attention_mask=attention_mask, ...
+        )
         for layer in self.layers:
-            hidden_states = layer(hidden_states, attention_mask, **kwargs)
+            hidden_states = layer(hidden_states, causal_mask, **kwargs)

Suppressing this rule

Add a # trf-ignore: TRF025 comment on the flagged line or the line directly above it. See Suppressing rules for whole-file directives and when a suppression is the wrong answer.