Update conditional logic and -> or in SAM postprocessing

[hwuebben]

should be ors here

[HuggingFaceDocBuilderDev]

The documentation is not available anymore as the PR was closed or merged.

[younesbelkada]

Thanks for the PR, I think that it should be as it is, input_points, input_labels and input_boxes are expected to be a list of lists (or a list of lists of lists), hence the and condition

[hwuebben]

@younesbelkada exactly but with the current implementation it could also be a list of arrays and the exception would not be thrown

[younesbelkada]

Ah yes, you are right, that sounds correct to me! Thanks for the heads up

[amyeroberts]

Thanks for the fix!

[ydshieh]

FYI, this PR triggered 3 test failures on the CI.

(The third one is GPU OOM, which is likely caused by the other 2 failures).

cc @younesbelkada

tests/models/sam/test_modeling_sam.py::SamModelIntegrationTest::test_inference_mask_generation_one_point_one_bb
(line 235)  ValueError: Input boxes must be a list of list of list of floating integers.
tests/models/sam/test_modeling_sam.py::SamModelIntegrationTest::test_inference_mask_generation_one_point_one_bb_zero
(line 235)  ValueError: Input boxes must be a list of list of list of floating integers.
tests/models/sam/test_modeling_sam.py::SamModelIntegrationTest::test_inference_mask_generation_two_points_batched
(line 808)  torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 2.00 GiB (GPU 0; 14.76 GiB total capacity; 11.65 GiB already allocated; 792.75 MiB free; 12.74 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation.  See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF

[karthikdatta98]

ValueError: Input boxes must be a list of list of list of floating integers. I AM Still getting this error, when even run the example notebooks, what do I do? I just have a list of bounding box coordinated(flatenned)

[younesbelkada]

Hi @karthikdatta98
thanks for reporting, in huggingface/notebooks#409 I modified the notebook accordingly to show how to correctly pass bounding boxes