paddle.nn.func.cosine_embedding_loss
#16728
Conversation
|
@hmahmood24 please review this PR |
ivy-leaves
added
the
Paddle Frontend
There was a problem hiding this comment.
Hey @mobley-trent
Sorry for being late to review this PR🙂, somehow didn't notice it.
Would you update with the latest master branch changes for me to be able to run tests locally.
Thanks😊
There was a problem hiding this comment.
Hey @mobley-trent
Below are comments for changes, also is it possible to check the loss for large dimensions like 5 😅, try to explore if that's possible.
Thanks😊
| ) | ||
|
|
||
| prod_sum = (input1 * input2).sum(axis=-1) | ||
| mag_square1 = ivy.square(input1).sum(axis=-1) + 10e-12 |
There was a problem hiding this comment.
For proper notation we can use
| mag_square1 = ivy.square(input1).sum(axis=-1) + 10e-12 | |
| mag_square1 = ivy.square(input1).sum(axis=-1) + 1e-12 |
There was a problem hiding this comment.
Won't this intefere with the final values ?
There was a problem hiding this comment.
It won’t interfere as this is epsilon value and the notation should be scientific😄
The function only works for 1D or 2D tensors 😅 @hirwa-nshuti |
There was a problem hiding this comment.
Hey @mobley-trent
Now the tests are passing and all looks great😊
Feel free to merge this item🙂
Closes #16727