Skip to content

[CI] Fix Fast GPU tests on PR #10912

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

Merged
merged 8 commits into from
Feb 27, 2025
Merged

[CI] Fix Fast GPU tests on PR #10912

Show file tree
Hide file tree
Changes from 5 commits
Commits
Show all changes
8 commits
Select commit Hold shift + click to select a range
e2d2650
update
DN6 Feb 25, 2025
d108c18
update
DN6 Feb 26, 2025
4756522
update
DN6 Feb 26, 2025
721501c
update
DN6 Feb 26, 2025
828dd32
Merge branch 'main' into fast-gpu-test-fixes
DN6 Feb 26, 2025
fac5514
Merge branch 'main' into fast-gpu-tests
sayakpaul Feb 27, 2025
cca8e14
update
DN6 Feb 27, 2025
d1fa030
Merge branch 'fast-gpu-tests' of https://github.com/huggingface/diffu…
DN6 Feb 27, 2025
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
9 changes: 4 additions & 5 deletions tests/models/test_modeling_common.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1169,17 +1169,16 @@ def test_disk_offload_without_safetensors(self):
base_output = model(**inputs_dict)

model_size = compute_module_sizes(model)[""]
max_size = int(self.model_split_percents[0] * model_size)
# Force disk offload by setting very small CPU memory
max_memory = {0: max_size, "cpu": int(0.1 * max_size)}

with tempfile.TemporaryDirectory() as tmp_dir:
model.cpu().save_pretrained(tmp_dir, safe_serialization=False)

with self.assertRaises(ValueError):
max_size = int(self.model_split_percents[0] * model_size)
max_memory = {0: max_size, "cpu": max_size}
# This errors out because it's missing an offload folder
new_model = self.model_class.from_pretrained(tmp_dir, device_map="auto", max_memory=max_memory)

max_size = int(self.model_split_percents[0] * model_size)
max_memory = {0: max_size, "cpu": max_size}
new_model = self.model_class.from_pretrained(
tmp_dir, device_map="auto", max_memory=max_memory, offload_folder=tmp_dir
)
Expand Down
5 changes: 3 additions & 2 deletions tests/models/transformers/test_models_transformer_omnigen.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -30,6 +30,7 @@ class OmniGenTransformerTests(ModelTesterMixin, unittest.TestCase):
model_class = OmniGenTransformer2DModel
main_input_name = "hidden_states"
uses_custom_attn_processor = True
model_split_percents = [0.1, 0.1, 0.1]

@property
def dummy_input(self):
Expand Down Expand Up @@ -73,9 +74,9 @@ def prepare_init_args_and_inputs_for_common(self):
"num_attention_heads": 4,
"num_key_value_heads": 4,
"intermediate_size": 32,
"num_layers": 1,
"num_layers": 20,
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Isn’t it a bit too much for CPUs?

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Overall model size is still very small (2.6 MB). Adds 2 seconds to the current CPU tests.

"pad_token_id": 0,
"vocab_size": 100,
"vocab_size": 1000,
"in_channels": 4,
"time_step_dim": 4,
"rope_scaling": {"long_factor": list(range(1, 3)), "short_factor": list(range(1, 3))},
Expand Down
6 changes: 4 additions & 2 deletions tests/models/transformers/test_models_transformer_sd3.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -33,6 +33,7 @@
class SD3TransformerTests(ModelTesterMixin, unittest.TestCase):
model_class = SD3Transformer2DModel
main_input_name = "hidden_states"
model_split_percents = [0.8, 0.8, 0.9]

@property
def dummy_input(self):
Expand Down Expand Up @@ -67,7 +68,7 @@ def prepare_init_args_and_inputs_for_common(self):
"sample_size": 32,
"patch_size": 1,
"in_channels": 4,
"num_layers": 1,
"num_layers": 4,
"attention_head_dim": 8,
"num_attention_heads": 4,
"caption_projection_dim": 32,
Expand Down Expand Up @@ -107,6 +108,7 @@ def test_gradient_checkpointing_is_applied(self):
class SD35TransformerTests(ModelTesterMixin, unittest.TestCase):
model_class = SD3Transformer2DModel
main_input_name = "hidden_states"
model_split_percents = [0.8, 0.8, 0.9]

@property
def dummy_input(self):
Expand Down Expand Up @@ -141,7 +143,7 @@ def prepare_init_args_and_inputs_for_common(self):
"sample_size": 32,
"patch_size": 1,
"in_channels": 4,
"num_layers": 2,
"num_layers": 4,
"attention_head_dim": 8,
"num_attention_heads": 4,
"caption_projection_dim": 32,
Expand Down