github/ColossalAI
1
0
Fork 0
mirror of https://github.com/hpcaitech/ColossalAI.git synced 2025-09-02 01:28:31 +00:00
Code Issues Packages Projects Releases Wiki Activity

[device] alpha beta profiler (#2311)

Browse Source 
* [device] alpha beta profiler

* add usage

* fix variable name
This commit is contained in:
YuliangLiu0306
2023-01-05 16:39:55 +08:00
committed by GitHub
parent f1bc2418c4
commit 9c9246c0d9
5 changed files with 227 additions and 129 deletions
Download Patch File Download Diff File Expand all files Collapse all files

31
tests/test_device/test_alpha_beta.py
View File

@@ -1,13 +1,32 @@
import pytest
from functools import partial
from colossalai.device import profile_alpha_beta
import pytest
import torch.multiprocessing as mp
from colossalai.device import AlphaBetaProfiler
from colossalai.initialize import launch
from colossalai.logging import disable_existing_loggers
from colossalai.testing import parameterize, rerun_if_address_is_in_use
from colossalai.utils import free_port
def check_alpha_beta(rank, physical_devices, world_size, port):
disable_existing_loggers()
launch(config={}, rank=rank, world_size=world_size, host='localhost', port=port, backend='nccl')
profiler = AlphaBetaProfiler(physical_devices)
ab_dict = profiler.profile_ab()
for _, (alpha, beta) in ab_dict.items():
assert alpha > 0 and alpha < 1e-4 and beta > 0 and beta < 1e-10
@pytest.mark.skip(reason="Skip because assertion fails for CI devices")
def test_profile_alpha_beta():
physical_devices = [0, 1, 2, 3]
(alpha, beta) = profile_alpha_beta(physical_devices)
assert alpha > 0 and alpha < 1e-4 and beta > 0 and beta < 1e-10
@pytest.mark.dist
@parameterize('physical_devices', [[0, 1, 2, 3], [0, 3]])
@rerun_if_address_is_in_use()
def test_profile_alpha_beta(physical_devices):
world_size = 4
run_func = partial(check_alpha_beta, physical_devices=physical_devices, world_size=world_size, port=free_port())
mp.spawn(run_func, nprocs=world_size)
if __name__ == '__main__':