[smaug]DeepLearning workload SMAUG result summary

[samialabed]

The current environment does not sum over all accelerators in the summary file.

the SMAUG environment for each layer it creates a different accelerator run, to properly support SMAUG we will need to address that and add the ability to sum over the SMAUG results.
yaoyuannnn/experiments#1 (comment)