The following images show the graphs of the models after conversion (click to enlarge):
|-
|4
|8084
|Interestingly, 7 actual processes are created beside the main one. Four of them, however, seem to be constantly in sleep state.
|}
The same tests were repeated using a network file system (NFS) over an Ethernet connection, too. No significant variations in the prediction times were observed.
In conclusion, to maximize the performance in terms of execution time, the model has to be fully-quantized and the number of threads has to be specified explicitly.