Checks / pattern
Attention without positional encoding
Check R10. Runs in the editor as you build, in CI through the
GitHub Action, and over the wire at
POST /api/v1/check. Milliseconds, before any GPU is billed.
warn
pattern
R10
| Trigger | Model contains attention layers but no positional encoding (sinusoidal, learned, RoPE, ALiBi). |
|---|---|
| Why | Self-attention is permutation-invariant. Without PE, the model is a bag-of-words and can't learn order. |
| Source | Vaswani et al. 2017, Attention Is All You Need, Section 3.5. |
Why it is not a lint you can ignore
A structural mistake does not fail at review time and it does not fail at import
time. It fails when the module is constructed on the training node, after the job was queued and
the dataset was downloaded. That is why this runs before the spend and not after it.
Run this check on your own model
Free, no account needed
Every check
41 structural checks: 6 guardrail gates and 35 architecture advisor rules. See the full catalogue.
← R09 Deep network with no residual connections ยท R11 Sigmoid / Tanh in deep networks →