fused_transpose_split_quant

paddle.incubate.nn.functional. fused_transpose_split_quant ( x, tokens_per_expert, pow_2_scales=False ) [source]