Read the equation
The sigmoid 1/(1+exp(−z)) is applied to each expert logit independently. It compresses each finite real logit into (0,1), without normalizing across experts. These values are routing affinities; they become relative mixture weights only after gathering selected entries and normalizing them.
Terms & definitions
| Term | Meaning in this source |
|---|---|
| logits | Raw expert logits |
| scores | Independent sigmoid affinities |
How it works
Read the definition from its right-hand expression to the quantity on the left. Its surrounding source records the intended input, state, or parameter context. Where the source leaves an operation unspecified, this page preserves that gap.
Assumptions & limits
Source recovery establishes that the expression appears in this record. Its mathematical validity, implementation behavior, and interpretation must be assessed separately. See the linked source and research notes for the checks available for this version.
Original context
Read the surrounding source passage
1. logits = F.linear( h.float32 , gate.weight.float32 , None ) -> (T, 256)
Both operands explicitly cast to float32 although gate.weight is BF16.
2. scores = sigmoid(logits) -> (T, 256)
Sigmoid, not softmax. Scores do not sum to 1 and are not a distribution.
3. scores_for_choice = scores + e_score_correction_bias -> (T, 256)
The F32 (256,) bias, broadcast over the token dimension.Exact original notation
scores = sigmoid(logits) -> (T, 256)