test_2d.py 2.04 KB
Newer Older
zbian's avatar
zbian committed
1
2
3
#!/usr/bin/env python
# -*- encoding: utf-8 -*-

4
5
from functools import partial

zbian's avatar
zbian committed
6
import pytest
Frank Lee's avatar
Frank Lee committed
7
8
import torch
import torch.multiprocessing as mp
zbian's avatar
zbian committed
9
from colossalai.core import global_context as gpc
アマデウス's avatar
アマデウス committed
10
from colossalai.initialize import launch
11
from colossalai.logging import disable_existing_loggers
12
13
from colossalai.utils import free_port

14
15
16
17
18
19
from checks_2d.check_layer_2d import (check_classifier_given_embed_weight, check_classifier_no_given_weight,
                                      check_embed, check_layernorm, check_linear, check_loss, check_patch_embed,
                                      check_vocab_parallel_classifier_given_embed_weight,
                                      check_vocab_parallel_classifier_no_given_weight, check_vocab_parallel_embed,
                                      check_vocab_parallel_loss)
from checks_2d.check_operation_2d import check_AB, check_ABT, check_ATB
zbian's avatar
zbian committed
20

21
CONFIG = dict(parallel=dict(pipeline=dict(size=1), tensor=dict(size=4, mode='2d')), )
zbian's avatar
zbian committed
22
23
24
25
26
27
28
29
30
31
32


def check_operations():
    check_AB()
    check_ABT()
    check_ATB()


def check_layer():
    check_linear()
    check_layernorm()
33
34
35
36
37
38
39
40
41
42
    check_embed()
    check_patch_embed()
    check_vocab_parallel_embed()
    check_classifier_no_given_weight()
    check_vocab_parallel_classifier_no_given_weight()
    check_classifier_given_embed_weight()
    check_vocab_parallel_classifier_given_embed_weight()
    check_loss()
    check_vocab_parallel_loss()

zbian's avatar
zbian committed
43

44
def check_layer_and_operation(rank, world_size, port):
45
46
    disable_existing_loggers()
    launch(config=CONFIG, rank=rank, world_size=world_size, host='localhost', port=port, backend='nccl')
Frank Lee's avatar
Frank Lee committed
47

48
49
50
    torch.backends.cuda.matmul.allow_tf32 = False
    torch.backends.cudnn.allow_tf32 = False
    torch.backends.cudnn.deterministic = True
アマデウス's avatar
アマデウス committed
51
    # check_operations()
zbian's avatar
zbian committed
52
53
    check_layer()
    gpc.destroy()
Frank Lee's avatar
Frank Lee committed
54
55
56
57
58
59
    torch.cuda.empty_cache()


@pytest.mark.dist
def test_2d():
    world_size = 4
60
    run_func = partial(check_layer_and_operation, world_size=world_size, port=free_port())
Frank Lee's avatar
Frank Lee committed
61
    mp.spawn(run_func, nprocs=world_size)
zbian's avatar
zbian committed
62
63
64
65


if __name__ == '__main__':
    test_2d()