Search Ctrl K
Home / Tools
THE FEUTEX TOOLKIT

Quantized Model Scale Overhead

Include per-group scale metadata in low-bit weight storage.

AI Infrastructure workspace● Runs in your browser
01 Your input
02 Your result
View or copy raw result data
FormulaPacked weights plus metadata = (params*1e6*bits/8+Math.ceil(params*1e6/group)*scale)/1073741824; Effective bits per weight = bits+8*scale/group

Made for real tasks

  • Include per-group scale metadata in low-bit weight storage.
Explore step-by-step workflows →

Worked example

Inputs

Quantized parameters
1000 millions
Bits per weight
4 bits
Weights per scale group
128 weights
Metadata bytes per group
4 bytes

Expected results

Packed weights plus metadata
0.4947651178 GiB
Effective bits per weight
4.25 bits

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 quantized parameters, bits per weight, weights per scale group, metadata bytes per group in the displayed units.
  2. Calculate the scenario, then compare it with measurements before changing a system.

Try this example

{
  "params": 1000,
  "bits": 4,
  "group": 128,
  "scale": 4
}

Accuracy and scope

Uniform groups and packed weights; alignment, unquantized tensors and runtime buffers excluded.

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.