Function diffusionLoss

diffusionLoss(
    model: RuntimeModel,
    promptIds: number[],
    targetIds: number[],
    tMin: number,
    tMax: number,
    vocab: number,
): MlxArray
Denoising cross-entropy on corrupted canvas positions (the reference diffusion_loss). Returns a scalar loss MlxArray, differentiable w.r.t. the mounted LoRA.
Parameters
- model: RuntimeModel
- promptIds: number[]
- targetIds: number[]
- tMin: number
- tMax: number
- vocab: number
Returns MlxArray
- Defined in packages/training/src/diffusion-lora.ts:89