TRF044

cache_position must not appear as a parameter in modeling code; it is removed framework surface.

Default	Enabled
Scope	All models
Source	mlinter/trf044.py
Show in terminal	mlinter --rule TRF044

What it does

Checks every function in modeling_*.py and modular_*.py for a parameter named cache_position.

Why is this bad?

cache_position was removed from all models in v5. Code that reintroduces it (usually copied from pre-v5 sources) threads a dead argument through every layer; the cache update call is past_key_values.update(key_states, value_states, self.layer_idx) with no position threading.

Example

 def forward(
     self,
     hidden_states,
     past_key_values=None,
-    cache_position=None,
     **kwargs,
 ):
-    key_states, value_states = past_key_values.update(key_states, value_states, self.layer_idx, cache_position)
+    key_states, value_states = past_key_values.update(key_states, value_states, self.layer_idx)

Suppressing this rule

Add a # trf-ignore: TRF044 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.