fix training error when training on GPU

[Abhijit-2592]

When training on GPU, the embedding layer throws the following error:

RuntimeError: Expected object of device type cuda but got device type cpu for argument # 3 'index' in call to _th_index_select

This is fixed by casting the CPU Tensors to CUDA Tensors. Also added the line which says to move the created conll files inside separate sub-directories inside data directory before parsing it with spaCy