LearningShapelets is much slower on GPU than CPU

I took following sample code from tslearn doc and executed on machine with GPU. The strange thing is, it's taking ~18min on GPU whereas code runs in few seconds on CPU.
Can you please tell whether the Shapelet Learning in tslearn has GPU support? If yes, do I need to explicitly specify that somewhere in source code. As I can see Matrix_Profile computation need implementation parameter to be specified as gpu_stump for GPU support. But I am not sure whether that is relate to LearningShapelets function or not.

Since I need to use it for large dataset, accelerating it with GPU is required for faster training.
Immediate help on this would be very much appreciated. Thanks

Start by reproducing the documented LearningShapelets example on the reported GPU and CPU setups, then inspect how LearningShapelets relates to Matrix_Profile and the gpu_stump implementation parameter. Confirm whether GPU execution is supported and identify the cause of the timing difference; done means the behavior and any required usage are documented or tracked with a reproducible fix.