TRF002

base_model_prefix should be a non-empty canonical string when defined on PreTrainedModel classes.

Default	Enabled
Scope	All models
Source	mlinter/trf002.py
Show in terminal	mlinter --rule TRF002

What it does

Checks that base_model_prefix, when set, is a non-empty, whitespace-free string literal.

Why is this bad?

Invalid prefixes can break weight loading key mapping and base model access patterns.

Example

 class AcmePreTrainedModel(PreTrainedModel):
-    base_model_prefix = ""
+    base_model_prefix = "model"

Suppressing this rule

Add a # trf-ignore: TRF002 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.

Allowlisted models

1 model are exempt from TRF002 in mlinter/rules.toml, because they predate the convention and cannot be changed without breaking backward compatibility.

Show the 1 allowlisted model

• lighton_ocr