View source on GitHub |
Gets the estimated training steps per epoch.
mediapipe_model_maker.model_util.get_steps_per_epoch(
steps_per_epoch: Optional[int] = None,
batch_size: Optional[int] = None,
train_data: Optional[mediapipe_model_maker.model_util.dataset.Dataset
] = None
) -> int
- If
steps_per_epoch
is set, returnssteps_per_epoch
directly. - Else if we can get the length of training data successfully, returns
train_data_length // batch_size
.
Args | |
---|---|
steps_per_epoch
|
int, training steps per epoch. |
batch_size
|
int, batch size. |
train_data
|
training data. |
Returns | |
---|---|
Estimated training steps per epoch. |
Raises | |
---|---|
ValueError
|
if both steps_per_epoch and train_data are not set. |