llama.cpp

convert : fix squeeze for ssm_conv tensors

#12573

Merged