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

[SAM] Fixes pipeline and adds a dummy pipeline test #23684

Merged
merged 2 commits into from
May 23, 2023
Merged

[SAM] Fixes pipeline and adds a dummy pipeline test #23684

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
d112bc5
add a dummy pipeline test
younesbelkada May 23, 2023
ab5244a
change test name
younesbelkada May 23, 2023
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/transformers/models/sam/image_processing_sam.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -934,7 +934,7 @@ def _generate_crop_boxes(
cropped_images, point_grid_per_crop = _generate_crop_images(
crop_boxes, image, points_grid, layer_idxs, target_size, original_size
)

crop_boxes = np.array(crop_boxes)
Copy link
Collaborator

Choose a reason for hiding this comment

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

when it currently fails, what is the type of crop_boxes? Is it a torch tensor? If so, we can't rely on np.array unless we are 100% sure crop_boxes is never on cuda.

Copy link
Collaborator

Choose a reason for hiding this comment

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

Pre-processing should never be done on cuda no?

Copy link
Collaborator

@ydshieh ydshieh May 23, 2023 •
edited
Loading

Choose a reason for hiding this comment

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

OK, my bad, didn't realize it's pre instead post processing. And #22970 actually removes the device stuff.

crop_boxes = crop_boxes.astype(np.float32)
points_per_crop = np.array([point_grid_per_crop])
points_per_crop = np.transpose(points_per_crop, axes=(0, 2, 1, 3))
Expand Down
8 changes: 7 additions & 1 deletion tests/models/sam/test_modeling_sam.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -20,7 +20,7 @@

import requests

from transformers import SamConfig, SamMaskDecoderConfig, SamPromptEncoderConfig, SamVisionConfig
from transformers import SamConfig, SamMaskDecoderConfig, SamPromptEncoderConfig, SamVisionConfig, pipeline
from transformers.testing_utils import require_torch, slow, torch_device
from transformers.utils import is_torch_available, is_vision_available

Expand Down Expand Up @@ -751,3 +751,9 @@ def test_inference_mask_generation_three_boxes_point_batch(self):
iou_scores = outputs.iou_scores.cpu()
self.assertTrue(iou_scores.shape == (1, 3, 3))
torch.testing.assert_allclose(iou_scores, EXPECTED_IOU, atol=1e-4, rtol=1e-4)

def test_dummy_pipeline_generation(self):
generator = pipeline("mask-generation", model="facebook/sam-vit-base", device=torch_device)
raw_image = prepare_image()

_ = generator(raw_image, points_per_batch=64)
Comment on lines +755 to +759
Copy link
Collaborator

Choose a reason for hiding this comment

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

I am surprised that this is not decorated by require_vision, but even more surprised that this test file has no require_vision at all, but having

if is_vision_available():
    from PIL import Image

This has to be fixed, but it's OK if we do this in a separate PR and maybe delegate the task to the original SAM modeling author in transformers.