Estimate the VRAM needed for model weights and inference from parameter count, precision bytes and activation overhead, to choose deployment hardware.
訓練還需最佳化器狀態(通常 2–4 倍權重)。