erelon

fuse_matmul_add_bias_into_gemm not working with batch size

2

Hi, When using fuse_matmul_add_bias_into_gemm I expect that even with batch size the layers will fuse. Apparently, this is not supported. I can't see what is the reason for this. If...