Method __init__

sat/dit_video_concat.py:547–561

(self, num_layers, in_features, hidden_features, bias=False)

class SwiGLUMixin(BaseMixin):
    def __init__(self, num_layers, in_features, hidden_features, bias=False):
        super().__init__()
        self.w2 = nn.ModuleList(
            [
                ColumnParallelLinear(
                    in_features,
                    hidden_features,
                    gather_output=False,
                    bias=bias,
                    module=self,
                    name="dense_h_to_4h_gate",
                )
                for i in range(num_layers)
            ]
        )

    def mlp_forward(self, hidden_states, **kw_args):
        x = hidden_states