omicverse.micro.plot_mmvec_training

omicverse.micro.plot_mmvec_training(mmvec, ax=None)

Training (and validation) loss curve for a fitted MMvec.

Plots mmvec.loss_history_ (and val_loss_history_ if a validation split was kept) per epoch. Marks the best-validation epoch with a vertical line — that’s where the saved checkpoint came from. A monotonically falling validation loss is healthy; a rising val loss while train keeps falling is overfitting.

参数:

• mmvec (MMvec) – Fitted model.

• ax (matplotlib axes, optional) – Axes to draw on; created if not provided.

返回类型:

matplotlib axes.