torch.js has not been released yet.

torch.autograd.BackwardOptions

export interface BackwardOptions {
  /**
   * If false, the graph used to compute the grads will be freed.
   * Setting to true allows computing higher order derivatives.
   * @default false
   */
  retain_graph?: boolean;

  /**
   * If true, graph of the derivative will be constructed, allowing
   * to compute higher order derivative products.
   * @default false
   */
  create_graph?: boolean;

  /**
   * Inputs w.r.t. which the gradient will be accumulated into .grad.
   * All other tensors will be ignored. If not provided, the gradient
   * is accumulated into all leaf tensors.
   */
  inputs?: Tensor | Tensor[];
}

retain_graph(boolean)optional: – If false, the graph used to compute the grads will be freed. Setting to true allows computing higher order derivatives.
create_graph(boolean)optional: – If true, graph of the derivative will be constructed, allowing to compute higher order derivative products.
inputs(Tensor | Tensor[])optional: – Inputs w.r.t. which the gradient will be accumulated into .grad. All other tensors will be ignored. If not provided, the gradient is accumulated into all leaf tensors.

Options for backward computation.