How this calculator works
Enter layer input dimension, layer output dimension, lora rank, adapted projections, adapted layers. Select Calculate to apply the displayed formula and review the labeled results.
Formula / method
LoRA parameters = rank × (input dimension + output dimension) × projections × layers
Worked example
Rank 16 across four 4,096×4,096 projections in 32 layers adds about 16.8M trainable parameters.
Assumptions and limitations
Each adapted projection uses one rank-down and one rank-up matrix.
Biases, embeddings, convolutional adapters, mixed dimensions and checkpoint metadata are excluded.