Description
Return aligned feed-forward expansion width.
Language-model bookkeeping separates scores, probabilities, dimensions, capacity and positional transforms. A local ranking, padding or coordinate operation can illustrate a model component without running a model, sampling a token, fitting calibration or demonstrating output quality.
Return aligned feed-forward expansion width.
When to use this tool
Use Transformer Ffn Expansion Width Factor when checking the stated model-dimension, ranking, probability-prefix or local transform convention.
How to use the tool and how it works
- Enter modelWidth, expansion, alignment. Use the units and conventions below.
- Run the tool. Correct any format, range, or cross-field input error before interpreting the output.
- Read rawWidth, alignedWidth, actualExpansion. Use the output convention below.
Return aligned feed-forward expansion width. Require positive whole modelWidth/alignment and expansion > 0; raw = width*expansion, aligned rounds upward to an alignment multiple, actualExpansion = aligned/modelWidth.1
| Input field | Meaning and units | Accepted input |
|---|---|---|
| modelWidth | Model Width | Required. Safe integer; minimum 1. |
| expansion | Expansion | Required. Finite number; strictly above 0. |
| alignment | Alignment | Required. Safe integer; minimum 1. |
| Output field | Interpretation | Returned shape |
|---|---|---|
| rawWidth | Raw Width | Required. Finite number. |
| alignedWidth | Aligned Width | Required. Safe integer. |
| actualExpansion | Actual Expansion | Required. Finite number. |
{
"input": {
"modelWidth": 128,
"expansion": 4,
"alignment": 64
},
"output": {
"rawWidth": 512,
"alignedWidth": 512,
"actualExpansion": 4
}
}Limitations and assumptions
- This is dimension bookkeeping, not a trained transformer, parameter count, memory estimate or recommended architecture.
- Inputs must satisfy both the field types and the cross-field conditions above. Invalid input causes an error; a successful false flag or documented placeholder is a different result.
- Numeric calculations use binary floating point unless exact decimal-integer arithmetic is explicitly stated. No hidden measurement tolerance or decimal rounding is added; nonfinite or out-of-contract results are rejected. Numeric signed-zero results are exported as zero.
Alternative or Complementary methods
Use a complete implementation with verified tensor/token conventions and appropriate model evaluation when actual inference, memory, sampling or quality results are required.
References
Similar or alternative tools
- Token Bucket Padding Efficiency Term
Return supplied token-bucket padding efficiency.
- Router Capacity Overflow Term
Return per-expert assigned-load overflow.