Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Update fsdp-tp checkpointing test #3603

Open
wants to merge 2 commits into
base: main
Choose a base branch
from
Open

Update fsdp-tp checkpointing test #3603

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
237e7b1
update test
eitanturok Sep 6, 2024
6116c21
only look at relevant tests
eitanturok Sep 6, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
44 changes: 22 additions & 22 deletions tests/trainer/test_fsdp_checkpoint.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -796,29 +796,29 @@ def mock_get_checkpoint_validation_function():


@pytest.mark.gpu
@pytest.mark.parametrize('use_remote', [pytest.param(True, marks=pytest.mark.remote), False])
# @pytest.mark.parametrize('use_remote', [pytest.param(True, marks=pytest.mark.remote), False])
@pytest.mark.parametrize(
'weights_only,optimizer,precision,autoresume,load_ignore_keys,use_symlink,use_tp,use_hsdp',
[
pytest.param(False, 'adamw', 'amp_bf16', False, None, False, False, False, marks=pytest.mark.world_size(2)),
pytest.param(True, 'adamw', 'amp_bf16', False, None, False, False, False, marks=pytest.mark.world_size(2)),
pytest.param(False, 'adam', 'amp_bf16', False, None, False, False, False, marks=pytest.mark.world_size(2)),
pytest.param(False, 'adamw', 'amp_fp16', False, None, False, False, False, marks=pytest.mark.world_size(2)),
pytest.param(False, 'adamw', 'amp_bf16', True, None, False, False, False, marks=pytest.mark.world_size(2)),
pytest.param(
False,
'adamw',
'amp_bf16',
False,
['rng'],
False,
False,
False,
marks=pytest.mark.world_size(2),
),
pytest.param(False, 'adamw', 'amp_bf16', False, None, True, False, False, marks=pytest.mark.world_size(2)),
# pytest.param(False, 'adamw', 'amp_bf16', False, None, False, False, False, marks=pytest.mark.world_size(2)),
# pytest.param(True, 'adamw', 'amp_bf16', False, None, False, False, False, marks=pytest.mark.world_size(2)),
# pytest.param(False, 'adam', 'amp_bf16', False, None, False, False, False, marks=pytest.mark.world_size(2)),
# pytest.param(False, 'adamw', 'amp_fp16', False, None, False, False, False, marks=pytest.mark.world_size(2)),
# pytest.param(False, 'adamw', 'amp_bf16', True, None, False, False, False, marks=pytest.mark.world_size(2)),
# pytest.param(
# False,
# 'adamw',
# 'amp_bf16',
# False,
# ['rng'],
# False,
# False,
# False,
# marks=pytest.mark.world_size(2),
# ),
# pytest.param(False, 'adamw', 'amp_bf16', False, None, True, False, False, marks=pytest.mark.world_size(2)),
pytest.param(False, 'adamw', 'amp_bf16', False, None, False, True, False, marks=pytest.mark.world_size(4)),
pytest.param(False, 'adamw', 'amp_bf16', False, None, False, False, True, marks=pytest.mark.world_size(4)),
# pytest.param(False, 'adamw', 'amp_bf16', False, None, False, False, True, marks=pytest.mark.world_size(4)),
],
)
@pytest.mark.filterwarnings(r'ignore:TypedStorage is deprecated.:UserWarning')
Expand All @@ -835,13 +835,13 @@ def test_fsdp_partitioned_state_dict_load(
use_symlink: bool,
use_tp: bool,
use_hsdp: bool,
use_remote,
s3_bucket,
s3_ephemeral_prefix,
request,
use_remote = False,
):
if use_tp:
pytest.skip('TP on PyTorch 2.3 has sharded state dict issues.')
if use_tp and version.parse(torch.__version__) < version.parse('2.4.0'):
pytest.skip('TP has sharded state dict issues before PyTorch 2.4.')
if weights_only and autoresume:
pytest.skip('Weights only with autoresume is not supported')
if (use_tp or use_hsdp) and version.parse(torch.__version__) < version.parse('2.3.0'):
Expand Down
Loading