Search Ctrl K
Home / Tools
THE FEUTEX TOOLKIT

LoRA Adapter Parameter Count

Count trainable low-rank matrices for repeated linear projections.

AI Infrastructure workspace● Runs in your browser
01 Your input
02 Your result
View or copy raw result data
FormulaAdapter parameters = rank*(input+output)*matrices; Adapter weight memory = rank*(input+output)*matrices*bytes/1048576
Sources and model references

Made for real tasks

  • Count trainable low-rank matrices for repeated linear projections.
Explore step-by-step workflows →

Worked example

Inputs

Input width
4096 dimensions
Output width
4096 dimensions
Adapter rank
16 rank
Adapted projections
64 matrices
Bytes per parameter
2 bytes

Expected results

Adapter parameters
8388608 parameters
Adapter weight memory
16 MiB

Load the example in the workspace to reproduce these results. Display rounding does not change the underlying calculation.

How to use this tool

  1. Enter input width, output width, adapter rank, adapted projections, bytes per parameter in the displayed units.
  2. Calculate the scenario, then compare it with measurements before changing a system.

Try this example

{
  "input": 4096,
  "output": 4096,
  "rank": 16,
  "matrices": 64,
  "bytes": 2
}

Accuracy and scope

Standard two-matrix LoRA without bias; excludes gradients, optimizer states and saved base weights.

Is my data uploaded?

This tool runs locally in your browser. Input and output are not sent to a processing server. Copy or download your result before leaving the page.