`ecnet.callbacks`¶

Training callbacks used by ECNet.fit.

Training callback objects/functions

class ecnet.callbacks.CallbackOperator[source]¶

Bases: object

CallbackOperator: executes individual callback steps at each step

add_cb(cb)[source]¶

on_train_begin()[source]¶

on_train_end()[source]¶

on_epoch_begin(epoch)[source]¶

on_epoch_end(epoch)[source]¶

on_batch_begin(batch)[source]¶

on_batch_end(batch)[source]¶

on_loss_begin(batch)[source]¶

on_loss_end(batch)[source]¶

on_step_begin(batch)[source]¶

on_step_end(batch)[source]¶

class ecnet.callbacks.Callback[source]¶

Bases: object

Base Callback object

on_train_begin()[source]¶

on_train_end()[source]¶

on_epoch_begin(epoch)[source]¶

on_epoch_end(epoch)[source]¶

on_batch_begin(batch)[source]¶

on_batch_end(batch)[source]¶

on_loss_begin(batch)[source]¶

on_loss_end(batch)[source]¶

on_step_begin(batch)[source]¶

on_step_end(batch)[source]¶

class ecnet.callbacks.LRDecayLinear(init_lr, decay_rate, optimizer)[source]¶

Bases: Callback

Parameters:

init_lr (float)
decay_rate (float)

on_epoch_begin(epoch)[source]¶

Return type:: bool
Parameters:: epoch (int)

Training halted if:: new learing rate == 0.0

class ecnet.callbacks.Validator(loader, model, eval_iter, patience)[source]¶

Bases: Callback

Parameters:

eval_iter (int)
patience (int)

on_epoch_end(epoch)[source]¶

Return type:: bool
Parameters:: epoch (int)

Training halted if:: number of epochs since last lowest valid. MAE > specified patience

on_train_end()[source]¶

After training, recall weights when lowest valid. MAE occurred

Return type:: bool