Hi guys, I followed the Harvard Annotated Transformer at Annotated Transformer and everything runs ok with text and integers. However, as the Transformer is an autoregressive model, I’d like to bypass the Embedding layer, given that it only accepts .long() data type (integers) and I have float data for Time Series forecasting. I searched for nn.Embedding source code in PyTorch but I guess it is written in C. How can I customize the nn.Embedding in a way it keeps the original input data as output or accepts floats as inputs without integer requirements/transformations?

Here is the class being used.

```
class Embeddings(nn.Module):
def __init__(self, d_model, vocab):
super(Embeddings, self).__init__()
self.lut = nn.Embedding(vocab, d_model) #here
self.d_model = d_model
def forward(self, x):
return self.lut(x) * math.sqrt(self.d_model)
```

Ahead of the code, we have:

```
def make_model(src_vocab, tgt_vocab, N=6,
d_model=512, d_ff=2048, h=8, dropout=0.1):
"Helper: Construct a model from hyperparameters."
c = copy.deepcopy
attn = MultiHeadedAttention(h, d_model)
ff = PositionwiseFeedForward(d_model, d_ff, dropout)
position = PositionalEncoding(d_model, dropout)
model = EncoderDecoder(
Encoder(EncoderLayer(d_model, c(attn), c(ff), dropout), N),
Decoder(DecoderLayer(d_model, c(attn), c(attn),
c(ff), dropout), N),
nn.Sequential(Embeddings(d_model, src_vocab), c(position)),
nn.Sequential(Embeddings(d_model, tgt_vocab), c(position)),
Generator(d_model, tgt_vocab))
for p in model.parameters():
if p.dim() > 1:
nn.init.xavier_uniform(p)
return model```
I tried to use `torch.cat` without success.
Thanks in advance
```