default_per_channel_weight_fake_quant

torch.ao.quantization.fake_quantize.default_per_channel_weight_fake_quant

Default fake_quant for per-channel weights. Observer is memoryless since averaging_constant is 1.