test_state_dict.py 2.05 KB
Newer Older
1
2
3
4
5
6
7
#!/usr/bin/env python
# -*- encoding: utf-8 -*-

from functools import partial

import pytest
import torch
8
9
10
from common import CONFIG

import colossalai
11
from colossalai.testing import parameterize, rerun_if_address_is_in_use, spawn
12
13
14
15
from colossalai.zero.legacy.init_ctx import ZeroInitContext
from colossalai.zero.legacy.shard_utils import BucketTensorShardStrategy, TensorShardStrategy
from colossalai.zero.legacy.sharded_model import ShardedModelV2
from colossalai.zero.legacy.sharded_model.utils import col_model_deepcopy
16
from tests.components_to_test.registry import non_distributed_component_funcs
17

18

19
@parameterize("shard_strategy_class", [TensorShardStrategy, BucketTensorShardStrategy])
20
def run_zero_state_dict(shard_strategy_class):
21
    test_models = ['repeated_computed_layers', 'resnet18']
22
    shard_strategy = shard_strategy_class()
23
24
    for model_name in test_models:
        get_components_func = non_distributed_component_funcs.get_callable(model_name)
25
        model_builder, train_dataloader, test_dataloader, optimizer, criterion = get_components_func()
26

27
        with ZeroInitContext(target_device=torch.device('cuda', torch.cuda.current_device()),
28
                             shard_strategy=shard_strategy,
29
                             shard_param=True):
30
31
32
33
34
35
36
            zero_model = model_builder(checkpoint=True)
        zero_model = ShardedModelV2(zero_model, shard_strategy)

        model = model_builder(checkpoint=True).half()
        col_model_deepcopy(zero_model, model)
        model = model.cuda()

37
38
        zero_state_dict = zero_model.state_dict()
        for key, val in model.state_dict().items():
39
            assert torch.equal(val, zero_state_dict[key].to(val.device))
40
41


42
43
44
45
46
def run_dist(rank, world_size, port):
    colossalai.launch(config=CONFIG, rank=rank, world_size=world_size, host='localhost', port=port, backend='nccl')
    run_zero_state_dict()


47
@pytest.mark.dist
ver217's avatar
ver217 committed
48
@pytest.mark.parametrize("world_size", [1, 2])
49
@rerun_if_address_is_in_use()
50
def test_zero_state_dict(world_size):
51
    spawn(run_dist, world_size)
52
53
54


if __name__ == '__main__':
55
    test_zero_state_dict(2)