Differences between torch.nn.TransformerDecoder and mindspore.nn.TransformerDecoder

torch.nn.TransformerDecoder

class torch.nn.TransformerDecoder(
    decoder_layer,
    num_layers,
    norm=None
)(tgt, memory, tgt_mask=None, memory_mask=None, tgt_key_padding_mask=None, memory_key_padding_mask=None)

For more information, see torch.nn.TransformerDecoder.

mindspore.nn.TransformerDecoder

class mindspore.nn.TransformerDecoder(
    decoder_layer,
    num_layers,
    norm=None
)(tgt, memory, tgt_mask=None, memory_mask=None, tgt_key_padding_mask=None, memory_key_padding_mask=None)

For more information, see mindspore.nn.TransformerDecoder.

Differences

The code implementation and parameter update logic of mindspore.nn.TransformerDecoder optimizer is mostly the same with torch.nn.TransformerDecoder.

Categories	Subcategories	PyTorch	MindSpore	Difference
Parameters	Parameter 1	decoder_layer	decoder_layer	Consistent function
	Parameter 2	num_layers	num_layers	Consistent function
	Parameter 3	norm	norm	Consistent function
Input	Input 1	tgt	tgt	Consistent function
	Input 2	memory	memory	Consistent function
	Input 3	tgt_mask	tgt_mask	In MindSpore, dtype can be set as float or bool Tensor; in Pytorch dtype can be set as float, byte or bool Tensor.
	Input 4	memory_mask	memory_mask	In MindSpore, dtype can be set as float or bool Tensor; in Pytorch dtype can be set as float, byte or bool Tensor.
	Input 5	tgt_key_padding_mask	tgt_key_padding_mask	In MindSpore, dtype can be set as float or bool Tensor; in Pytorch dtype can be set as byte or bool Tensor.
	Input 6	memory_key_padding_mask	memory_key_padding_mask	In MindSpore, dtype can be set as float or bool Tensor; in Pytorch dtype can be set as byte or bool Tensor.

Code Example

# PyTorch
import torch
from torch import nn

decoder_layer = nn.TransformerDecoderLayer(d_model=512, nhead=8)
transformer_decoder = nn.TransformerDecoder(decoder_layer, num_layers=6)
memory = torch.rand(10, 32, 512)
tgt = torch.rand(20, 32, 512)
out = transformer_decoder(tgt, memory)
print(out.shape)
#torch.Size([20, 32, 512])

# MindSpore
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)