mindspore.nn.TransformerDecoder

View Source On Gitee
class mindspore.nn.TransformerDecoder(decoder_layer, num_layers, norm=None)[source]

Transformer Decoder module with multi-layer stacked of TransformerDecoderLayer, including multihead self attention, cross attention and feedforward layer.

Parameters
Inputs:
  • tgt (Tensor): The sequence to the decoder. For unbatched input, the shape is \((T, E)\) ; otherwise if batch_first=False in TransformerDecoderLayer, the shape is \((T, N, E)\) and if batch_first=True , the shape is \((T, N, E)\), where \((T)\) is the target sequence length. Supported types: float16, float32, float64.

  • memory (Tensor): The sequence from the last layer of the encoder. Supported types: float16, float32, float64.

  • tgt_mask (Tensor, optional): the mask of the tgt sequence. The shape is \((T, T)\) or \((N*nhead, T, T)\) , where nhead is the arguent in TransformerDecoderLayer. Supported types: float16, float32, float64, bool. Default: None.

  • memory_mask (Tensor, optional): the mask of the memory sequence. The shape is \((T, S)\) . Supported types: float16, float32, float64, bool. Default: None.

  • tgt_key_padding_mask (Tensor, optional): the mask of the tgt keys per batch. Supported types: float16, float32, float64, bool. Default: None.

  • memory_key_padding_mask (Tensor, optional): the mask of the memory keys per batch. The shape is \((S)\) for unbatched input, otherwise \((N, S)\) . Supported types: float16, float32, float64, bool. Default: None.

Outputs:

Tensor. The shape and dtype of Tensor is the same with tgt .

Supported Platforms:

Ascend GPU CPU

Examples

>>> import mindspore as ms
>>> import numpy as np
>>> decoder_layer = ms.nn.TransformerDecoderLayer(d_model=512, nhead=8)
>>> transformer_decoder = ms.nn.TransformerDecoder(decoder_layer, num_layers=6)
>>> memory = ms.Tensor(np.random.rand(10, 32, 512), ms.float32)
>>> tgt = ms.Tensor(np.random.rand(20, 32, 512), ms.float32)
>>> out = transformer_decoder(tgt, memory)
>>> print(out.shape)
(20, 32, 512)