timeseries: benchmark gpu

[laszewsk]

• [ ] we need to be able to set if GPU is used or not so we can measure effect of benchmark
• [x] ource.directory = "/N/u/ketanp/cm/cloudmesh-timeseries/notebook/benchmarks/../../data" is hardcoded, must be relative

https://github.com/cloudmesh/cloudmesh-timeseries/blob/master/notebook/benchmarks/ok-v2-lstm-predict-n-new.ipynb

[kpimparkar]

• If we use tensorflow in the script to be run then we can add tf.debugging.set_log_device_placement(True) in the script which indicates if any task was run on a gpu or cpu. The output of the script can be saved in a logfile to analyze if GPU was used. Example: pytest test_03_job_gpu.py

[kpimparkar]

The hardcoded path of data is removed. The relative path is added.