Closed weiyanlin117 closed 1 year ago
Implement with added tests. Aligned with layer_norm_cpu_ref.c variable name Compare to Pytorch implementation, removed redundant computation(reuse) for ahp[x] = (ap1[x] - meanp2[0]) * inv_stdp2[0];
Implement with added tests. Aligned with layer_norm_cpu_ref.c variable name Compare to Pytorch implementation, removed redundant computation(reuse) for ahp[x] = (ap1[x] - meanp2[0]) * inv_stdp2[0];