fix a bug of sequence_slice layer when batch_size=1.

hidden capi symbols

Fix pre-commit-config.yaml

Cluster train doc update

fix ids reverse order in fillGenOutputs.

Add a lookup table op and a CUDA helper.

Make cudnn convolution layer and projection support for dilation

init mnist

register rowwise add gpu kernel

master client: retry connecting to etcd

Feature/minus op

reduce some compile warning

Add a sequence slice layer.