Transformer XL
Model from Keita Kurita.
- class flood_forecast.transformer_xl.transformer_xl.MultiHeadAttention(d_input: int, d_inner: int, n_heads: int = 4, dropout: float = 0.1, dropouta: float = 0.0)[source]
- __init__(d_input: int, d_inner: int, n_heads: int = 4, dropout: float = 0.1, dropouta: float = 0.0)[source]
Initialize internal Module state, shared by both nn.Module and ScriptModule.
- forward(input_: FloatTensor, pos_embs: FloatTensor, memory: FloatTensor, u: FloatTensor, v: FloatTensor, mask: FloatTensor | None = None)[source]
- pos_embs: we pass the positional embeddings in separately
because we need to handle relative positions
input shape: (seq, bs, self.d_input) pos_embs shape: (seq + prev_seq, bs, self.d_input) output shape: (seq, bs, self.d_input)
- class flood_forecast.transformer_xl.transformer_xl.PositionwiseFF(d_input, d_inner, dropout)[source]
- __init__(d_input, d_inner, dropout)[source]
Initialize internal Module state, shared by both nn.Module and ScriptModule.
- forward(input_: FloatTensor) FloatTensor [source]
Define the computation performed at every call.
Should be overridden by all subclasses.
Note
Although the recipe for forward pass needs to be defined within this function, one should call the
Module
instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.
- class flood_forecast.transformer_xl.transformer_xl.DecoderBlock(n_heads, d_input, d_head_inner, d_ff_inner, dropout, dropouta=0.0)[source]
- __init__(n_heads, d_input, d_head_inner, d_ff_inner, dropout, dropouta=0.0)[source]
Initialize internal Module state, shared by both nn.Module and ScriptModule.
- forward(input_: FloatTensor, pos_embs: FloatTensor, u: FloatTensor, v: FloatTensor, mask=None, mems=None)[source]
Define the computation performed at every call.
Should be overridden by all subclasses.
Note
Although the recipe for forward pass needs to be defined within this function, one should call the
Module
instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.
- class flood_forecast.transformer_xl.transformer_xl.PositionalEmbedding(d)[source]
-
- forward(positions: LongTensor)[source]
Define the computation performed at every call.
Should be overridden by all subclasses.
Note
Although the recipe for forward pass needs to be defined within this function, one should call the
Module
instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.
- class flood_forecast.transformer_xl.transformer_xl.StandardWordEmbedding(num_embeddings, embedding_dim, div_val=1, sample_softmax=False)[source]
- __init__(num_embeddings, embedding_dim, div_val=1, sample_softmax=False)[source]
Initialize internal Module state, shared by both nn.Module and ScriptModule.
- forward(input_: LongTensor)[source]
Define the computation performed at every call.
Should be overridden by all subclasses.
Note
Although the recipe for forward pass needs to be defined within this function, one should call the
Module
instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.
- class flood_forecast.transformer_xl.transformer_xl.TransformerXL(num_embeddings, n_layers, n_heads, d_model, d_head_inner, d_ff_inner, dropout=0.1, dropouta=0.0, seq_len: int = 0, mem_len: int = 0)[source]
- __init__(num_embeddings, n_layers, n_heads, d_model, d_head_inner, d_ff_inner, dropout=0.1, dropouta=0.0, seq_len: int = 0, mem_len: int = 0)[source]
Initialize internal Module state, shared by both nn.Module and ScriptModule.
- forward(idxs: LongTensor, target: LongTensor, memory: List[FloatTensor] | None = None) Dict[str, Tensor] [source]
Define the computation performed at every call.
Should be overridden by all subclasses.
Note
Although the recipe for forward pass needs to be defined within this function, one should call the
Module
instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.