# GPT-J
# Input token size = 2016
# Output token size = 32
# Batch Size = 12
# Num Beams = 4
# M = num_beams * batch_size = 48

--reset
--dir=FWD_I
--dt=f16

mb96768ic4096oc4096_n"gptj:gemm0*112"
mb96768ic4096oc16384_n"gptj:gemm3*28"
mb96768ic16384oc4096_n"gptj:gemm4*28"
mb96768ic4096oc50400_n"gptj:gemm5*1"
mb48ic4096oc4096_n"gptj:gemm6*3472"
mb48ic4096oc16384_n"gptj:gemm9*868"
mb48ic16384oc4096_n"gptj:gemm10*868"
mb48ic4096oc50400_n"gptj:gemm11*31"
