Method __init__

openrec/modeling/decoders/cppd_decoder.py:129–164

(
        self,
        dim,
        num_heads,
        mlp_ratio=4.0,
        qkv_bias=False,
        qk_scale=None,
        drop=0.0,
        attn_drop=0.0,
        drop_path=0.0,
        act_layer=nn.GELU,
        norm_layer=nn.LayerNorm,
        epsilon=1e-6,
    )

class DecoderLayer(nn.Module):

    def __init__(
        self,
        dim,
        num_heads,
        mlp_ratio=4.0,
        qkv_bias=False,
        qk_scale=None,
        drop=0.0,
        attn_drop=0.0,
        drop_path=0.0,
        act_layer=nn.GELU,
        norm_layer=nn.LayerNorm,
        epsilon=1e-6,
    ):
        super().__init__()
        self.norm1 = norm_layer(dim, eps=epsilon)
        self.mixer = Attention(
            dim,
            num_heads=num_heads,
            qkv_bias=qkv_bias,
            qk_scale=qk_scale,
            attn_drop=attn_drop,
            proj_drop=drop,
        )

        # NOTE: drop path for stochastic depth, we shall see if this is better than dropout here
        self.drop_path = DropPath(drop_path) if drop_path > 0.0 else Identity()
        self.norm2 = norm_layer(dim, eps=epsilon)
        mlp_hidden_dim = int(dim * mlp_ratio)
        self.mlp_ratio = mlp_ratio
        self.mlp = Mlp(
            in_features=dim,
            hidden_features=mlp_hidden_dim,
            act_layer=act_layer,
            drop=drop,
        )

    def forward(self, q, kv, key_mask=None):
        x1 = self.norm1(q + self.drop_path(self.mixer(q, kv, key_mask)))