MetaExecutor Class

[jan-janssen]

The MetaExecutor manages multiple Executors in the background.

Example:

from time import sleep
from pympipool.external_interfaces.meta import MetaExecutor

def calc(i):
    sleep(5)
    return i

with MetaExecutor(max_workers=2) as exe:
    fs_1 = exe.submit(calc, 1)
    fs_2 = exe.submit(calc, 2)
    print(fs_1.result(), fs_2.result())

[jan-janssen]

GPU example test.py:

from pympipool import MetaExecutor
from tensorflow.python.client import device_lib

def get_available_gpus():
    local_device_protos = device_lib.list_local_devices()
    return [(x.name, x.physical_device_desc) for x in local_device_protos if x.device_type == 'GPU']


if __name__ == "__main__":
    with MetaExecutor(max_workers=4, cores_per_worker=1, gpus_per_worker=1, enable_flux_backend=True) as exe:
        fs_1 = exe.submit(get_available_gpus)
        fs_2 = exe.submit(get_available_gpus)
        fs_3 = exe.submit(get_available_gpus)
        fs_4 = exe.submit(get_available_gpus)
    
    print(fs_1.result())
    print(fs_2.result())
    print(fs_3.result())
    print(fs_4.result())

Output:

>>> python test.py
2023-07-27 14:00:45.008247: I tensorflow/core/platform/cpu_feature_guard.cc:182] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.
To enable the following instructions: SSE4.1 SSE4.2 AVX AVX2 FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.
[('/device:GPU:0', 'device: 0, name: NVIDIA A100-SXM4-40GB, pci bus id: 0000:01:00.0, compute capability: 8.0')]
[('/device:GPU:0', 'device: 0, name: NVIDIA A100-SXM4-40GB, pci bus id: 0000:41:00.0, compute capability: 8.0')]
[('/device:GPU:0', 'device: 0, name: NVIDIA A100-SXM4-40GB, pci bus id: 0000:c1:00.0, compute capability: 8.0')]
[('/device:GPU:0', 'device: 0, name: NVIDIA A100-SXM4-40GB, pci bus id: 0000:81:00.0, compute capability: 8.0')]

[jan-janssen]

The same works on GPU clusters:

import socket
from pympipool import MetaExecutor
from tensorflow.python.client import device_lib

def get_available_gpus():
    local_device_protos = device_lib.list_local_devices()
    return [(x.name, x.physical_device_desc, socket.gethostname()) for x in local_device_protos if x.device_type == 'GPU']


if __name__ == "__main__":
    with MetaExecutor(max_workers=2, cores_per_worker=1, gpus_per_worker=1, enable_flux_backend=True) as exe:
        fs_1 = exe.submit(get_available_gpus)
        fs_2 = exe.submit(get_available_gpus)
    
    print(fs_1.result())
    print(fs_2.result())

Output:

[('/device:GPU:0', 'device: 0, name: Tesla V100S-PCIE-32GB, pci bus id: 0000:84:00.0, compute capability: 7.0', 'cn138')]
[('/device:GPU:0', 'device: 0, name: Tesla V100S-PCIE-32GB, pci bus id: 0000:84:00.0, compute capability: 7.0', 'cn139')]