Model Training and Testing

Fit

class flexflow.core.flexflow_cffi.FFModel

fit(x=None, y=None, batch_size=None, epochs=1)

Trains the model for a fixed number of epochs (iterations on a dataset).

Parameters

x (Dataloader) – Input data. It can be a Dataloader instance or a list of Dataloader instances.
y (Dataloader) – Target data (label). It can be a Dataloader instance or a list of Dataloader instances.
batch_size (int) – Number of samples per gradient update. It must be identical with -b or --batch-size from the command line.
epochs (int) – Number of epochs to train the model. An epoch is an iteration over the entire x and y data provided. The default value is 1.

Returns

None – no returns.

class flexflow.core.flexflow_cffi.FFModel

eval(x=None, y=None, batch_size=None)

Returns the loss value & metrics values for the model in test mode.

Parameters

x (Dataloader) – Input data. It can be a Dataloader instance or a list of Dataloader instances.
y (Dataloader) – Target data (label). It can be a Dataloader instance or a list of Dataloader instances.
batch_size (int) – Number of samples per gradient update. It must be identical with -b or --batch-size from the command line.
epochs (int) – Number of epochs to train the model. An epoch is an iteration over the entire x and y data provided. The default value is 1.

Returns

None – no returns.

class flexflow.core.flexflow_cffi.FFModel

backward(seq_length=None)

Backward propagation of all layers.

compute_metrics()

Compute performance metrics.

forward(seq_length=None)

Forward propagation of all layers.

reset_metrics()

Reset performance metrics.

update()

Update weights and biases of all layers.

zero_gradients()

Empty the gradients of all layers.