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

Adds better Affine support for GPUs when using CUDA 11. Introduces a new bias addition kernel for CUDA < 11 #778

Merged
merged 23 commits into from
Apr 9, 2021
Merged

Adds better Affine support for GPUs when using CUDA 11. Introduces a new bias addition kernel for CUDA < 11 #778

merged 23 commits into from
Apr 9, 2021

Commits on Dec 8, 2020

  1. Added new inference operator that peforms bias addition and optionall… 

    …y RELU for inference. When upgrading to cuda 11, the bias and relu can be fused into the matrix multiply with cublasLt.
    @rhenry-nv
    rhenry-nv committed Dec 8, 2020
    Configuration menu
    Copy the full SHA
    66006ec View commit details
    Browse the repository at this point in the history

  2. Fix compilation for CPU only. Missed DISPATCH10 call when CUDA was un… 

    …defined
    @rhenry-nv
    rhenry-nv committed Dec 8, 2020
    Configuration menu
    Copy the full SHA
    23a3223 View commit details
    Browse the repository at this point in the history

Commits on Dec 9, 2020

  1. Places fused affine under one node

    @rhenry-nv
    rhenry-nv committed Dec 9, 2020
    Configuration menu
    Copy the full SHA
    dbb1653 View commit details
    Browse the repository at this point in the history

  2. Adds basic affine support for cublaslt in cuda 11

    @rhenry-nv
    rhenry-nv committed Dec 9, 2020
    Configuration menu
    Copy the full SHA
    3e94764 View commit details
    Browse the repository at this point in the history

  3. Works around some bugs in CUDA 11

    @rhenry-nv
    rhenry-nv committed Dec 9, 2020
    Configuration menu
    Copy the full SHA
    4b41b86 View commit details
    Browse the repository at this point in the history

Commits on Dec 15, 2020

  1. Fixes API and removes comment

    @rhenry-nv
    rhenry-nv committed Dec 15, 2020
    Configuration menu
    Copy the full SHA
    cc30a32 View commit details
    Browse the repository at this point in the history

  2. Removes SPDX identifier in operator tests

    @rhenry-nv
    rhenry-nv committed Dec 15, 2020
    Configuration menu
    Copy the full SHA
    792669f View commit details
    Browse the repository at this point in the history

  3. Adds SPDX identifier in operator tests - will remove in a different PR

    @rhenry-nv
    rhenry-nv committed Dec 15, 2020
    Configuration menu
    Copy the full SHA
    0c0b8ac View commit details
    Browse the repository at this point in the history

  4. Removes NVIDIA notices

    @rhenry-nv
    rhenry-nv committed Dec 15, 2020
    Configuration menu
    Copy the full SHA
    6258fd5 View commit details
    Browse the repository at this point in the history

  5. Fixes windows compile errors

    @rhenry-nv
    rhenry-nv committed Dec 15, 2020
    Configuration menu
    Copy the full SHA
    764d7b8 View commit details
    Browse the repository at this point in the history

  6. Format changes

    @rhenry-nv
    rhenry-nv committed Dec 15, 2020
    Configuration menu
    Copy the full SHA
    95fb427 View commit details
    Browse the repository at this point in the history

  7. updates changelog

    @rhenry-nv
    rhenry-nv committed Dec 15, 2020
    Configuration menu
    Copy the full SHA
    4bbf17f View commit details
    Browse the repository at this point in the history

  8. Adds bigobj flag to windows builds

    @rhenry-nv
    rhenry-nv committed Dec 15, 2020
    Configuration menu
    Copy the full SHA
    256e397 View commit details
    Browse the repository at this point in the history

Commits on Mar 6, 2021

  1. Merge remote-tracking branch 'public/master' into cuda_11

    @rhenry-nv
    rhenry-nv committed Mar 6, 2021
    Configuration menu
    Copy the full SHA
    493c213 View commit details
    Browse the repository at this point in the history

Commits on Mar 23, 2021

  1. Merge branch 'master' of https://github.com/marian-nmt/marian-dev int… 

    …o cuda_11
    @rhenry-nv
    rhenry-nv committed Mar 23, 2021
    Configuration menu
    Copy the full SHA
    e5b549d View commit details
    Browse the repository at this point in the history

Commits on Mar 25, 2021

  1. refactor

    @emjotde
    emjotde committed Mar 25, 2021
    Configuration menu
    Copy the full SHA
    a484893 View commit details
    Browse the repository at this point in the history

  2. remove previous code

    @emjotde
    emjotde committed Mar 25, 2021
    Configuration menu
    Copy the full SHA
    9de84d4 View commit details
    Browse the repository at this point in the history

  3. add unit tests

    @emjotde
    emjotde committed Mar 25, 2021
    Configuration menu
    Copy the full SHA
    43c54ce View commit details
    Browse the repository at this point in the history

  4. fix incorrect function signature for CUDA 10 and smaller

    @emjotde
    emjotde committed Mar 25, 2021
    Configuration menu
    Copy the full SHA
    04b0f95 View commit details
    Browse the repository at this point in the history

  5. only use AffineWithReluNodeOp for float types

    @emjotde
    emjotde committed Mar 25, 2021
    Configuration menu
    Copy the full SHA
    5c7b6ee View commit details
    Browse the repository at this point in the history

  6. switch back to using affineWithRelu on gpu only for now

    @emjotde
    emjotde committed Mar 25, 2021
    Configuration menu
    Copy the full SHA
    c027c6d View commit details
    Browse the repository at this point in the history

Commits on Apr 2, 2021

  1. Merge pull request #2 from marian-nmt/mjd/refactor_cuda11 

    Small refactoring of cuda_11
    @rhenry-nv
    rhenry-nv authored Apr 2, 2021
    Configuration menu
    Copy the full SHA
    7ea2d09 View commit details
    Browse the repository at this point in the history

Commits on Apr 9, 2021

  1. Merge branch 'master' into cuda_11

    @emjotde
    emjotde authored Apr 9, 2021
    Configuration menu
    Copy the full SHA
    36d8f0c View commit details
    Browse the repository at this point in the history