[caffe2] make order btw div and mul in adgrad consistent #32974
Closed
Conversation
This file contains hidden or 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
Summary: Re-attempt of D18805426 . Decided to be consistent with PyTorch Adagrad There was an inconsistency in the order of operation between scalar and SIMD code when we compute Adagrad. This diff make them consistent by doing w += lr * grad / (sqrt(moment) + epsilon) in Adagrad and w += lr / (sqrt(moment) + epsilon) * grad in RowWiseSparseAdagrad. The Adagrad order is consistent with PyTorch (see aten/src/ATen/native/cpu/PointwiseOpsKernel.cpp addcmul_cpu_kernel function). The RowWiseSparseAdagrad order is to make compute more efficient. In RowWiseSparseAdagrad, lr / (sqrt(moment) + epsilon) is shared among all elements in the row And, we're not going to use FMA to be consistent with PyTorch (even though it provides a little accuracy benefit) Test Plan: CI Differential Revision: D19342865 fbshipit-source-id: 7189e7a8ba6c17773fb1ff99a78906eb8441c13b
|
This pull request was exported from Phabricator. Differential Revision: D19342865 |
facebook-github-bot
pushed a commit
that referenced
this pull request
Feb 17, 2020
Summary: Pull Request resolved: #32974 Pull Request resolved: pytorch/FBGEMM#286 Re-attempt of D18805426 . Decided to be consistent with PyTorch Adagrad There was an inconsistency in the order of operation between scalar and SIMD code when we compute Adagrad. This diff make them consistent by doing w += lr * grad / (sqrt(moment) + epsilon) in Adagrad and w += lr / (sqrt(moment) + epsilon) * grad in RowWiseSparseAdagrad. The Adagrad order is consistent with PyTorch (see aten/src/ATen/native/cpu/PointwiseOpsKernel.cpp addcmul_cpu_kernel function). The RowWiseSparseAdagrad order is to make compute more efficient. In RowWiseSparseAdagrad, lr / (sqrt(moment) + epsilon) is shared among all elements in the row And, we're not going to use FMA to be consistent with PyTorch (even though it provides a little accuracy benefit) Test Plan: CI Reviewed By: wx1988 Differential Revision: D19342865 fbshipit-source-id: e950c16f2e1c4a2f2a3ef53b1705db373c67f341
ttumiel
pushed a commit
to ttumiel/pytorch
that referenced
this pull request
Mar 4, 2020
Summary: Pull Request resolved: pytorch#32974 Pull Request resolved: pytorch/FBGEMM#286 Re-attempt of D18805426 . Decided to be consistent with PyTorch Adagrad There was an inconsistency in the order of operation between scalar and SIMD code when we compute Adagrad. This diff make them consistent by doing w += lr * grad / (sqrt(moment) + epsilon) in Adagrad and w += lr / (sqrt(moment) + epsilon) * grad in RowWiseSparseAdagrad. The Adagrad order is consistent with PyTorch (see aten/src/ATen/native/cpu/PointwiseOpsKernel.cpp addcmul_cpu_kernel function). The RowWiseSparseAdagrad order is to make compute more efficient. In RowWiseSparseAdagrad, lr / (sqrt(moment) + epsilon) is shared among all elements in the row And, we're not going to use FMA to be consistent with PyTorch (even though it provides a little accuracy benefit) Test Plan: CI Reviewed By: wx1988 Differential Revision: D19342865 fbshipit-source-id: e950c16f2e1c4a2f2a3ef53b1705db373c67f341
jspark1105
added a commit
to pytorch/FBGEMM
that referenced
this pull request
Mar 21, 2020
Summary: Pull Request resolved: pytorch/pytorch#32974 Pull Request resolved: #286 Re-attempt of D18805426 . Decided to be consistent with PyTorch Adagrad There was an inconsistency in the order of operation between scalar and SIMD code when we compute Adagrad. This diff make them consistent by doing w += lr * grad / (sqrt(moment) + epsilon) in Adagrad and w += lr / (sqrt(moment) + epsilon) * grad in RowWiseSparseAdagrad. The Adagrad order is consistent with PyTorch (see aten/src/ATen/native/cpu/PointwiseOpsKernel.cpp addcmul_cpu_kernel function). The RowWiseSparseAdagrad order is to make compute more efficient. In RowWiseSparseAdagrad, lr / (sqrt(moment) + epsilon) is shared among all elements in the row And, we're not going to use FMA to be consistent with PyTorch (even though it provides a little accuracy benefit) Reviewed By: wx1988 Differential Revision: D19342865 fbshipit-source-id: e950c16f2e1c4a2f2a3ef53b1705db373c67f341
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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.
Summary:
Re-attempt of D18805426 . Decided to be consistent with PyTorch Adagrad
There was an inconsistency in the order of operation between scalar and SIMD code when we compute Adagrad. This diff make them consistent by doing w += lr * grad / (sqrt(moment) + epsilon) in Adagrad and w += lr / (sqrt(moment) + epsilon) * grad in RowWiseSparseAdagrad.
The Adagrad order is consistent with PyTorch (see aten/src/ATen/native/cpu/PointwiseOpsKernel.cpp addcmul_cpu_kernel function). The RowWiseSparseAdagrad order is to make compute more efficient. In RowWiseSparseAdagrad, lr / (sqrt(moment) + epsilon) is shared among all elements in the row
And, we're not going to use FMA to be consistent with PyTorch (even though it provides a little accuracy benefit)
Test Plan: CI
Differential Revision: D19342865