NetworkOutputLayer

class NetworkOutputLayer.OutputLayer(loss, y, dtype=None, reshape_target=False, copy_input=None, copy_output=None, time_limit=0, use_source_index=False, auto_fix_target_length=False, sigmoid_outputs=False, exp_outputs=False, gauss_outputs=False, activation=None, prior_scale=0.0, log_prior=None, use_label_priors=0, compute_priors_via_baum_welch=False, compute_priors=False, compute_priors_exp_average=0, compute_priors_accumulate_batches=None, compute_distortions=False, softmax_smoothing=1.0, grad_clip_z=None, grad_discard_out_of_bound_z=None, normalize_length=False, exclude_labels=[], apply_softmax=True, batchwise_softmax=False, substract_prior_from_output=False, input_output_similarity=None, input_output_similarity_scale=1, scale_by_error=False, copy_weights=False, **kwargs)[source]
Parameters:
  • index (theano.Variable) – index for batches
  • loss (str) – e.g. ‘ce’
layer_class = 'softmax'[source]
create_bias(n, prefix='b', name='')[source]
entropy()[source]
Return type:theano.Variable
errors()[source]
Return type:theano.Variable
class NetworkOutputLayer.FramewiseOutputLayer(loss, y, dtype=None, reshape_target=False, copy_input=None, copy_output=None, time_limit=0, use_source_index=False, auto_fix_target_length=False, sigmoid_outputs=False, exp_outputs=False, gauss_outputs=False, activation=None, prior_scale=0.0, log_prior=None, use_label_priors=0, compute_priors_via_baum_welch=False, compute_priors=False, compute_priors_exp_average=0, compute_priors_accumulate_batches=None, compute_distortions=False, softmax_smoothing=1.0, grad_clip_z=None, grad_discard_out_of_bound_z=None, normalize_length=False, exclude_labels=[], apply_softmax=True, batchwise_softmax=False, substract_prior_from_output=False, input_output_similarity=None, input_output_similarity_scale=1, scale_by_error=False, copy_weights=False, **kwargs)[source]
Parameters:
  • index (theano.Variable) – index for batches
  • loss (str) – e.g. ‘ce’
cost()[source]
Return type:(theano.Variable | None, dict[theano.Variable,theano.Variable] | None)
Returns:cost, known_grads
cost_scale()[source]
class NetworkOutputLayer.DecoderOutputLayer(**kwargs)[source]
cost()[source]
class NetworkOutputLayer.SequenceOutputLayer(ce_smoothing=0.0, ce_target_layer_align=None, am_scale=1, gamma=1, bw_norm_class_avg=False, fast_bw_opts=None, seg_fast_bw_opts=None, loss_like_ce=False, trained_softmax_prior=False, sprint_opts=None, warp_ctc_lib=None, **kwargs)[source]
index_for_ctc()[source]
output_index()[source]
cost()[source]
Parameters:y – shape (time*batch,) -> label
Returns:error scalar, known_grads dict
errors()[source]
class NetworkOutputLayer.UnsupervisedOutputLayer(base, momentum=0.1, oracle=False, msteps=100, esteps=200, **kwargs)[source]
cost()[source]
errors()[source]
Return type:theano.Variable