VRAM–005VRAM Calculator
Runs entirely in your browser — nothing is uploaded.
Inputs
Architecture auto-fills from the preset
Same as attention heads if not using GQA.
Typical LoRA setups train roughly 0.1–2% of total parameters, depending on rank and target modules.
Output
| GPU | VRAM | GPUs needed |
|---|
| Precision | Weights | Total VRAM | Fits current GPU config |
|---|
Total VRAM
0 GB
Weights
0 GB
KV cache
0 GB
Activations
0 GB
Available
0 GB
Note
—
