Universal functions #29
Open
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
govereau
requested changes
Jan 21, 2025
| binop a x y (fun x y => .ok (x * y)) | ||
|
|
||
| def div (a : Type) [Div a] [Element a] (x y : Tensor) : Err Tensor := | ||
| binop a x y (fun x y => .ok (x / y)) |
There was a problem hiding this comment.
Should this check for divide by zero?
There was a problem hiding this comment.
Numpy just gives inf or nan in this case. E.g.
#np.full(5, 1, dtype='int8') / np.arange(5)
<ipython-input-26-4ea893b50512>:1: RuntimeWarning: divide by zero encountered in divide
np.full(5, 1, dtype='int8') / np.arange(5)
array([ inf, 1. , 0.5 , 0.33333333, 0.25 ])
Seems like the correct behavior to me. E.g. if we fail on nans we will have quite a different behavior than the current numpy and neuron compiler.
TensorLib/Ufunc.lean
Outdated
| return Element.arrayScalar acc | ||
|
|
||
| -- Sum with a single axis. | ||
| def sum1 (a : Type) [Add a] [Zero a] [Element a] (arr : Tensor) (axis : Nat) : Err Tensor := do |
There was a problem hiding this comment.
This came out pretty nice. Seems like the design is holding up well.
|
|
||
| private def uniq [BEq a] (xs : List a) : Bool := match xs with | ||
| | [] | [_] => true | ||
| | x1 :: x2 :: xs => x1 != x2 && uniq (x2 :: xs) |
There was a problem hiding this comment.
Hmm, I don't understand this function.
uniq [1,2,1] == True ?
Maybe, the input is sorted?
There was a problem hiding this comment.
Good catch, fixed.
| | .none => sum0 a arr | ||
| | .some axes => | ||
| if !(uniq axes) then .error "Duplicate axis elements" else | ||
| let axes := (List.mergeSort axes).reverse |
There was a problem hiding this comment.
Ah here is the sort, but maybe in the wrong place?
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
NumPy uses an abstraction of functions called Ufunc. It has a lot of widgets in NumPy, but for now we just support lifting element-wise operations and broadcasting.