espnet2.mt.frontend.embedding.PatchEmbedding

source

class espnet2.mt.frontend.embedding.PatchEmbedding(input_size: int = 400, embed_dim: int = 400, token_per_frame: int = 1, pos_enc_class=<class 'espnet.nets.pytorch_backend.transformer.embedding.PositionalEncoding'>, positional_dropout_rate: float = 0.1)

Bases: AbsFrontend

Embedding Frontend for text based inputs.

Initialize.

• Parameters:
  • input_size – Number of input tokens.
  • embed_dim – Embedding Size.
  • token_per_frame – number of tokens per frame in the input
  • pos_enc_class – PositionalEncoding or ScaledPositionalEncoding
  • positional_dropout_rate – dropout rate after adding positional encoding

forward(input: Tensor, input_lengths: Tensor) → Tuple[Tensor, Tensor]

Apply a sliding window on the input.

• Parameters:
  • input – Input (B, T)
  • input_lengths – Input lengths within batch.
• Returns: Output with dimensions (B, T // token_per_frame, D). Tensor: Output lengths within batch, devided by token_per_frame
• Return type: Tensor

output_size() → int

Return output length of feature dimension D, i.e. the embedding dim.