optimum/amd/brevitas/accelerate_utils.py (3 lines):
	- line 54: # TODO: Why no no_split_module_classes, clean_result parameters?
	- line 411: # TODO: Fix directly in accelerate.
	- line 422: # TODO: Fix this bug directly in accelerate. https://github.com/huggingface/accelerate/pull/2214 would fix the bug for RAM offliading.


optimum/amd/brevitas/quantizer.py (2 lines):
	- line 81: # TODO: fix
	- line 253: TODO: test this, and maybe use it by default?


optimum/amd/ryzenai/models/image_transforms.py (1 line):
	- line 38: # TODO: Replace the cv2 resize with PIL based resize after accuracy tests


optimum/amd/cli.py (1 line):
	- line 53: # TODO: how to handle NUMA nodes and socket affinity?


setup.py (1 line):
	- line 19: # TODO: unpin pytest once https://github.com/huggingface/transformers/pull/29154 is merged & released