conftest.py 1.86 KB
Newer Older
Tom Birch's avatar
Tom Birch committed
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
# Copyright (c) Facebook, Inc. and its affiliates. All rights reserved.
#
# This source code is licensed under the BSD license found in the
# LICENSE file in the root directory of this source tree.

# Copyright 2019 Kakao Brain
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#   http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

import functools
import os

import pytest
import torch

from fairscale.nn.model_parallel import destroy_model_parallel


@pytest.fixture(autouse=True)
def manual_seed_zero():
    torch.manual_seed(0)


def cuda_sleep_impl(seconds, cycles_per_ms):
    torch.cuda._sleep(int(seconds * cycles_per_ms * 1000))


@pytest.fixture(scope="session")
def cuda_sleep():
    # Warm-up CUDA.
    torch.empty(1, device="cuda")

    # From test/test_cuda.py in PyTorch.
    start = torch.cuda.Event(enable_timing=True)
    end = torch.cuda.Event(enable_timing=True)
    start.record()
    torch.cuda._sleep(1000000)
    end.record()
    end.synchronize()
    cycles_per_ms = 1000000 / start.elapsed_time(end)

    return functools.partial(cuda_sleep_impl, cycles_per_ms=cycles_per_ms)


def pytest_report_header():
    return f"torch: {torch.__version__}"


def pytest_runtest_setup(item):
    print(f"setup mpi function called")


def pytest_runtest_teardown(item):
    if "OMPI_COMM_WORLD_RANK" in os.environ:
        destroy_model_parallel()
        if torch.distributed.is_initialized():
            torch.distributed.destroy_process_group()