Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Matmul/Conv2d/Pool2d cuda kernels #333

Closed
Tracked by #278
coreylowman opened this issue Jan 7, 2023 · 3 comments · Fixed by #384
Closed
Tracked by #278

Matmul/Conv2d/Pool2d cuda kernels #333

coreylowman opened this issue Jan 7, 2023 · 3 comments · Fixed by #384
Assignees
@coreylowman
Labels
gpu Related to GPU support

Comments

@coreylowman
Copy link
Owner

These will be more challenging to implement, since they will rely on a lot of code and invoking cublas.

Also need to see if there's a way to simplify the implementation (potentially by having a device level matmul trait that matmul/conv2d can use)
@coreylowman coreylowman added the gpu Related to GPU support label Jan 7, 2023
@coreylowman coreylowman self-assigned this Jan 7, 2023
@coreylowman coreylowman mentioned this issue Jan 7, 2023
Closed
47 tasks
@coreylowman
Copy link
Owner Author

matmul tests blocked by #335

@coreylowman coreylowman mentioned this issue Jan 7, 2023
Merged
@coreylowman
Copy link
Owner Author

This will require sub views of CudaSlices, which will be addressed in cudarc

@M1ngXU
Copy link
Contributor

M1ngXU commented Jan 9, 2023

all of them except matmul exist in cudnn

coreylowman added a commit that referenced this issue Jan 16, 2023
@coreylowman
#333 conv2d forward implemented
dc21b98
This was referenced Jan 16, 2023
Merged
Merged
coreylowman added a commit that referenced this issue Jan 20, 2023
@coreylowman
Add cuda kernels for conv2d (#369)
b9617b1 
* #333 conv2d forward implemented

* Adding todos

* Fixing bug with conv2d cpu kernel

* Working commit of cuda kernel. backwards not passing for weight grad

* Some tests passing

* Cleanup, tests still failing

* Fixing bug with matmul with 1 dimensions
@coreylowman coreylowman mentioned this issue Jan 20, 2023
Merged
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees

@coreylowman coreylowman

Labels
gpu Related to GPU support
Projects
None yet
Milestone
No milestone
Development

Successfully merging a pull request may close this issue.

Rework pool2d internals & add pool2d cuda kernels coreylowman/dfdx
2 participants
@coreylowman @M1ngXU