Is the implementation of SinusoidalPositionEmbeddings in diffusion tutorial correct ?

```
class SinusoidalPositionEmbeddings(nn.Module):
def __init__(self, dim):
super().__init__()
self.dim = dim

def forward(self, time):
device = time.device
half_dim = self.dim // 2
embeddings = math.log(10000) / (half_dim - 1)
embeddings = torch.exp(torch.arange(half_dim, device=device) * -embeddings)
embeddings = time[:, None] * embeddings[None, :]
embeddings = torch.cat((embeddings.sin(), embeddings.cos()), dim=-1)
return embeddings
```
In the tutorial : https://github.com/huggingface/blog/blob/main/annotated-diffusion.md

for a given time-step, will this implementation not place all the sines first, and all the cosines later. (due to the line `embeddings = torch.cat((embeddings.sin(), embeddings.cos()), dim=-1)`)

should it not be alternating sines and cosines like `sin , cos, sin, cos ... and so on` ?

Start with the SinusoidalPositionEmbeddings implementation in annotated-diffusion.md and inspect how its output is consumed in the diffusion tutorial. Verify whether the sine/cosine ordering matches the intended positional embedding convention, then document the conclusion or update the tutorial if the implementation is incorrect.

Written by the indexing model from the issue text.