NVIDIA / NVIDIA/cutlass

[QST] Modify how to load Activations and Filters

Open
#1,950 2 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

? - Needs Triage inactive-30d inactive-90d question
Dominant language
C++
Stars
10.5k
Forks
2.1k
Avg merge
3d 11h
Merged PRs (30d)
7

Description

What is your question?
Hello, I’ve read the post about how convolution is implemented in CUTLASS. My idea is to create a custom convolution kernel, not by modifying the operation itself, but by applying transformations to the filters before operating.

My goal would be to do this when loading the filters and make it so in a parallel way. Is this possible in CUTLASS? Can it be achieved using an example from the /examples directory, or would internal code modifications be required?

Contributor guide

No contributing guide indexed for this repository

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start with the implicit GEMM convolution documentation and the linked conv2d_fprop_filter_tile_access_iterator_optimized.h implementation. Review the /examples directory to determine whether filter transformations during loading are supported there; done means establishing whether an example can demonstrate this or whether internal CUTLASS changes are required.

Written by the indexing model from the issue text.

Assessment

Tech stack
cpp
Domain
machine-learning, performance
Issue type
Feature
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.