Skip to content
This repository has been archived by the owner on Nov 22, 2022. It is now read-only.

Implement Blockwise Sparsification #1050

Closed

Conversation

psuzhanhy
Copy link
Contributor

Summary:
Support row or column blockwise sparsification. Group parameters into blocks and either an entire block is removed or retained.

  1. refactor sparsifier.py under optimizer/sparsifers and create "sparsifiers_lib". Reason: some sparsifiers depend on "models_lib" to extract particualr layers to perform sparsification. Yet "models_lib depends on "optimizers_lib". Refactoring prevents circular dependency.

  2. add blockwise_sparsifier.py and unit-tests.

Differential Revision: D17910541

@facebook-github-bot facebook-github-bot added the CLA Signed Do not delete this pull request or issue due to inactivity. label Oct 14, 2019
@facebook-github-bot
Copy link
Contributor

This pull request was exported from Phabricator. Differential Revision: D17910541

psuzhanhy added a commit to psuzhanhy/pytext that referenced this pull request Oct 15, 2019
Summary:
Pull Request resolved: facebookresearch#1050

Support row or column blockwise sparsification. Group parameters into blocks and either an entire block is removed or retained.

1. refactor sparsifier.py under optimizer/sparsifers and create "sparsifiers_lib". Reason: some sparsifiers depend on "models_lib" to extract particualr layers to perform sparsification. Yet "models_lib depends on "optimizers_lib". Refactoring prevents circular dependency.

2. add blockwise_sparsifier.py and unit-tests.

Differential Revision: D17910541

fbshipit-source-id: 81a35ed6de6ff5f97366cbb0a231424e9a7b0579
@facebook-github-bot
Copy link
Contributor

This pull request was exported from Phabricator. Differential Revision: D17910541

psuzhanhy added a commit to psuzhanhy/pytext that referenced this pull request Oct 15, 2019
Summary:
Pull Request resolved: facebookresearch#1050

Support row or column blockwise sparsification. Group parameters into blocks and either an entire block is removed or retained.

1. refactor sparsifier.py under optimizer/sparsifers and create "sparsifiers_lib". Reason: some sparsifiers depend on "models_lib" to extract particualr layers to perform sparsification. Yet "models_lib depends on "optimizers_lib". Refactoring prevents circular dependency.

2. add blockwise_sparsifier.py and unit-tests.

Differential Revision: D17910541

fbshipit-source-id: 1721ca8f324a8ff03c53b162b726133f3f1c22a6
@facebook-github-bot
Copy link
Contributor

This pull request was exported from Phabricator. Differential Revision: D17910541

1 similar comment
@facebook-github-bot
Copy link
Contributor

This pull request was exported from Phabricator. Differential Revision: D17910541

psuzhanhy added a commit to psuzhanhy/pytext that referenced this pull request Oct 21, 2019
Summary:
Pull Request resolved: facebookresearch#1050

Support row or column blockwise sparsification. Group parameters into blocks and either an entire block is removed or retained.

1. refactor sparsifier.py under optimizer/sparsifers and create "sparsifiers_lib". Reason: some sparsifiers depend on "models_lib" to extract particualr layers to perform sparsification. Yet "models_lib depends on "optimizers_lib". Refactoring prevents circular dependency.

2. add blockwise_sparsifier.py and unit-tests.

Differential Revision: D17910541

fbshipit-source-id: 841e36e1ab3699e6606e89240e7797e4e1ea0544
psuzhanhy added a commit to psuzhanhy/pytext that referenced this pull request Oct 21, 2019
Summary:
Pull Request resolved: facebookresearch#1050

Support row or column blockwise sparsification. Group parameters into blocks and either an entire block is removed or retained.

1. refactor sparsifier.py under optimizer/sparsifers and create "sparsifiers_lib". Reason: some sparsifiers depend on "models_lib" to extract particualr layers to perform sparsification. Yet "models_lib depends on "optimizers_lib". Refactoring prevents circular dependency.

2. add blockwise_sparsifier.py and unit-tests.

Differential Revision: D17910541

fbshipit-source-id: a86b53d70aebbca7fa4eb12aeada5d5b4297fb38
@facebook-github-bot
Copy link
Contributor

This pull request was exported from Phabricator. Differential Revision: D17910541

Summary:
Pull Request resolved: facebookresearch#1050

Support row or column blockwise sparsification. Group parameters into blocks and either an entire block is removed or retained.

1. refactor sparsifier.py under optimizer/sparsifers and create "sparsifiers_lib". Reason: some sparsifiers depend on "models_lib" to extract particualr layers to perform sparsification. Yet "models_lib depends on "optimizers_lib". Refactoring prevents circular dependency.

2. add blockwise_sparsifier.py and unit-tests.

Reviewed By: arbabu123

Differential Revision: D17910541

fbshipit-source-id: 7451beac9f974d7fa52595339438b73259f3139d
@facebook-github-bot
Copy link
Contributor

This pull request was exported from Phabricator. Differential Revision: D17910541

@facebook-github-bot
Copy link
Contributor

This pull request has been merged in 0e6132c.

Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
CLA Signed Do not delete this pull request or issue due to inactivity. Merged
Projects
None yet
Development

Successfully merging this pull request may close these issues.

None yet

2 participants