quantlab.model.predefined.transformer.PositionalEncoding# class quantlab.model.predefined.transformer.PositionalEncoding(d_model, max_len=1000)[source]# Bases: Module Qlib’s sinusoidal position encoding of time-major [N, B, D] input. Examples >>> PositionalEncoding(d_model=4)(torch.zeros(3, 2, 4))[1, 0] tensor([0.8415, 0.5403, 0.0100, 0.9999]) forward(x)[source]# Return x plus the encoding of its N steps.
