transformers

433d2a23 - Update SAM/SAM HQ attention implementation + fix Cuda sync issues (#39386)

Update SAM/SAM HQ attention implementation + fix Cuda sync issues (#39386) * update attention implementation and improve inference speed * modular sam_hq + fix integration tests on A10 * fixup * fix after review * softmax in correct place * return attn_weights in sam/sam_hq