M-Labs/nac3
7
1
Fork 18
Code Issues 93 Pull Requests 15 Releases Wiki Activity

ndstrides: [11] Implement general matmul & np_dot() #521

Closed
lyken wants to merge 2 commits from ndstrides-11-matmul into ndstrides-10-ops
pull from: ndstrides-11-matmul
merge into: M-Labs:ndstrides-10-ops
Conversation 2 Commits 2 Files Changed 14 +473 -412
lyken commented 2024-08-28 14:14:38 +08:00
Contributor
Copy Link
  1. Reimplements np_dot. There is no improvement on the types of ndarrays np_dot can handle. It is still limited to 1D 'dot' 1D and scalar 'dot' scalar.
  2. Adds general matrix multiplication.
    • 1D @ 2D, 2D @ 1D, and stacking are now supported.
    • Solves #397 (ndarray 1D matrix multiplication).

Other notes

On matmul's signature: Is the typechecker type correct? I have also encountered the same issue like np_size() when the other_ty of impl_matmul is simply set to [ndarray_unsized_t] - primitives.ndarray's dtype and ndims are substituted permanently. See commit "core/ndstrides: implement general matmul".

1. Reimplements `np_dot`. There is ***no*** improvement on the types of ndarrays `np_dot` can handle. It is still limited to 1D 'dot' 1D and scalar 'dot' scalar. 2. Adds general matrix multiplication. - 1D @ 2D, 2D @ 1D, and stacking are now supported. - Solves https://git.m-labs.hk/M-Labs/nac3/issues/397 (ndarray 1D matrix multiplication). ### Other notes On matmul's signature: Is the typechecker type correct? I have also encountered the same issue like `np_size()` when the `other_ty` of `impl_matmul` is simply set to `[ndarray_unsized_t]` - `primitives.ndarray`'s `dtype` and `ndims` are substituted permanently. See commit "core/ndstrides: implement general matmul".
lyken added 1 commit 2024-08-28 14:14:38 +08:00
core/ndstrides: implement general matmul b416ece921
lyken added 1 commit 2024-08-28 14:15:51 +08:00
core/ndstrides: implement np_dot() for scalars and 1D 75b2e80418
lyken added a new dependency 2024-08-28 14:21:29 +08:00
#522 ndstrides: [12] Reimplement builtins
lyken added a new dependency 2024-08-28 14:21:39 +08:00
#520 ndstrides: [10] Reimplement binops, unary ops, and cmpops.
lyken changed title from ndstrides: [11] Implement general matmul to ndstrides: [11] Implement general matmul & `np_dot()` 2024-08-30 15:00:18 +08:00
lyken force-pushed ndstrides-11-matmul from 75b2e80418 to 4baf1c64ed 2024-08-30 15:08:04 +08:00 Compare
lyken commented 2024-08-30 15:08:09 +08:00
Author
Contributor
Copy Link

Rebased.

Rebased.
derppening force-pushed ndstrides-11-matmul from 4baf1c64ed to e7aa2688e6 2024-10-18 14:33:31 +08:00 Compare
derppening force-pushed ndstrides-11-matmul from e7aa2688e6 to 7a3e34eb7a 2024-10-18 14:41:02 +08:00 Compare
derppening force-pushed ndstrides-11-matmul from 7a3e34eb7a to 693b7f3774 2024-10-18 15:01:47 +08:00 Compare
derppening commented 2025-01-06 16:51:20 +08:00
Collaborator
Copy Link

Superseded by #568.

Superseded by #568.
derppening closed this pull request 2025-01-06 16:51:21 +08:00

Pull request closed

This pull request cannot be reopened because the branch was deleted.
Sign in to join this conversation.
Reviewers
No Reviewers
Labels
Clear labels
future-extension
high-priority
low-priority
UROP0622
No Label
Milestone
No items
No Milestone
Assignees
Clear assignees
derppening ivan jordens LPTK mwojcik oarcher occheung rclovis sb10q srenblad
No Assignees
2 Participants
Notifications
Due Date
No due date set.
Blocks
#522 ndstrides: [12] Reimplement builtins
M-Labs/nac3
Depends on
#520 ndstrides: [10] Reimplement binops, unary ops, and cmpops.
M-Labs/nac3
Reference: M-Labs/nac3#521
Delete Branch "ndstrides-11-matmul"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?