TRF024

Layer dimensions must come from the config, not from an integer literal in the modeling file.

Default	Enabled
Scope	Models added on or after 2026-06-20
Source	mlinter/trf024.py
Show in terminal	mlinter --rule TRF024

What it does

In modeling_*.py and modular_*.py, checks torch.nn layer constructors (Linear, Embedding, LayerNorm, RMSNorm, GroupNorm, BatchNorm*, InstanceNorm*, Conv*d, ConvTranspose*d, Bilinear, MultiheadAttention) for an integer literal greater than 8 in a dimension position, whether passed positionally or by keyword (in_features, out_features, in_channels, out_channels, num_embeddings, embedding_dim, embed_dim, normalized_shape, num_channels, hidden_size). Operator-shape arguments such as kernel_size, stride, padding and num_groups are ignored, and literals up to 8 are allowed so scalar heads, binary classifiers and RGB channel counts stay clean. Models contributed before cutoff_date are exempt.

Why is this bad?

A hardcoded width silently pins the module to one checkpoint size: the same architecture at another scale loads with a shape mismatch, and from_pretrained cannot report which value is wrong because there is no config field to point at. It also splits the source of truth, so a config that is edited no longer describes the model that gets built.

Example

 class AcmeAtomEmbedding(nn.Module):
     def __init__(self, config):
         super().__init__()
-        self.proj = nn.Linear(768, 3072, bias=False)
-        self.norm = nn.LayerNorm(3072)
+        self.proj = nn.Linear(config.hidden_size, config.intermediate_size, bias=False)
+        self.norm = nn.LayerNorm(config.intermediate_size)

Suppressing this rule

Add a # trf-ignore: TRF024 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.