es-mpi

Utilities to run deep speech as a REST API and some MPI tests.

Run deep speech server

Install the requirements and run python deepspeech-server.py. API client code snippet is available in deepspeech-client.py.

Running MPI tests:

To know more about what MPI is, see here. First run the script install-mpi4py.sh to install all dependencies (tested only for mac). Then run mpi-test.py with the following command: mpirun -np 8 python mpi-test.py.

The np flag denotes the number of parallel workers (1 master and 7 slaves). If you want to change the number of workers, change the value of np and change the num_workers variable inside the code (both should have the same value)

If everything goes well, you should get an output like this:

Total time elapsed for slave 1: 0.28195691108703613
Total time elapsed for slave 2: 0.33805084228515625
Total time elapsed for slave 3: 0.3919670581817627
Total time elapsed for slave 4: 0.44113707542419434
Total time elapsed for slave 5: 0.48983216285705566
Total time elapsed for slave 6: 0.5376040935516357
Total time elapsed for slave 7: 0.589820146560669
Final result shape: (7, 1000000)
Total time elapsed for master 0: 1.1907389163970947