b2083
17c97fb0 · CUDA: mul_mat_vec_q max. batch size 8 -> 4 (#5370) · Feb 06, 2024