Skip to content

Fix model offloading and training tests + prevent examples timeout #14091

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

Open
GiGiKoneti wants to merge 2 commits into
base: main
Choose a base branch
from
+10 −9
Open

Fix model offloading and training tests + prevent examples timeout #14091

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
f16aa73
Fix AutoencoderVidTok output format and skip training tests on unsupp…
GiGiKoneti Jun 29, 2026
a6b117b
Merge branch 'main' into fix/vae-and-examples-ci-failures
GiGiKoneti Jul 1, 2026
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
2 changes: 1 addition & 1 deletion src/diffusers/models/autoencoders/autoencoder_vidtok.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1502,5 +1502,5 @@ def forward(
dec = dec[:, :, :-time_padding, :, :]

if not return_dict:
return dec
return (dec,)
return DecoderOutput(sample=dec)
2 changes: 0 additions & 2 deletions tests/models/autoencoders/test_models_autoencoder_vidtok.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -13,7 +13,6 @@
# See the License for the specific language governing permissions and
# limitations under the License.

import pytest
import torch

from diffusers import AutoencoderVidTok
Expand Down Expand Up @@ -80,7 +79,6 @@ def get_dummy_inputs(self) -> dict:


class TestAutoencoderVidTok(AutoencoderVidTokTesterConfig, ModelTesterMixin):
@pytest.mark.skip("VidTok output structure not compatible with recursive output check.")
Comment thread
GiGiKoneti marked this conversation as resolved.
def test_outputs_equivalence(self):
super().test_outputs_equivalence()

Expand Down
15 changes: 9 additions & 6 deletions tests/models/testing_utils/training.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -210,11 +210,14 @@ def test_mixed_precision_training(self):

# Test with bfloat16
if torch.device(torch_device).type != "cpu":
model.zero_grad()
with torch.amp.autocast(device_type=torch.device(torch_device).type, dtype=torch.bfloat16):
output = model(**inputs_dict, return_dict=False)[0]
if torch.device(torch_device).type == "cuda" and not torch.cuda.is_bf16_supported():
pytest.skip("bfloat16 training is not supported on this GPU.")
else:
model.zero_grad()
with torch.amp.autocast(device_type=torch.device(torch_device).type, dtype=torch.bfloat16):
output = model(**inputs_dict, return_dict=False)[0]

noise = torch.randn((output.shape[0],) + self.output_shape).to(torch_device)
loss = torch.nn.functional.mse_loss(output, noise)
noise = torch.randn((output.shape[0],) + self.output_shape).to(torch_device)
loss = torch.nn.functional.mse_loss(output, noise)

loss.backward()
loss.backward()
Loading