1x1 version of the standard convolution block.

This function is taken from the original tf repo. It ensures that all layers have a channel number that is divisible by 8 It can be seen

This function is taken from the original tf repo. It ensures that all layers have a channel number that is divisible by 8 It can be seen

Round weighted channel number (make divisible operation). Parameters: ---------- channels : int or float Original number of

This function is taken from the original tf repo. It ensures that all layers have a channel number that is divisible by 8 It can be seen

This function is taken from the original tf repo. It ensures that all layers have a channel number that is divisible by 8 It can be seen

Record line profiling information for the given Python function.

Calculate TF-same like padding size. Parameters: ---------- x : tensor Input tensor. kernel_size : int Convoluti

1x1 convolution

3x3 convolution with padding

Compute Receiver operating characteristic (ROC) with a score and label vector.

Naive forward of arcface. Note: Refer to https://github.com/ronghuaiyang/arcface-pytorch/blob/master/models/metrics.py#L35.

3x3 convolution with padding

3x3 version of the standard convolution block.

Depthwise version of the standard convolution block.

1x1 version of the pre-activated convolution block.

Print the stat of a registered function

Print the stat of each functions

Register the trace_callback only on demand

Show results for a single function.

Update status of lr. Args: iteration(int, optional): now training iteration of all epochs. Normally need not to s

Parallel forward in column dimension.