Skip to content

Conversation

@antoine-tran
Copy link

Summary:
The FBNet V2 consists of several basic blocks of pattern conv(2d) + BatchNorm + ReLU. These operators can be fused into one operator, greatly speeding up the computation in GPU.

This diff / PR adds option to build an FBNet V2 backbone in fused manner, i.e. it looks up recursively the blocks and sub-blocks for the known patterns, and fuse them using mobile_cv.arch.utils.fuse_utils

Differential Revision: D47395999

@facebook-github-bot facebook-github-bot added CLA Signed Do not delete this pull request or issue due to inactivity. fb-exported labels Jul 13, 2023
@facebook-github-bot
Copy link
Contributor

This pull request was exported from Phabricator. Differential Revision: D47395999

antoine-tran added a commit to antoine-tran/mobile-vision that referenced this pull request Jul 13, 2023
Summary:
X-link: facebookresearch/d2go#592

Pull Request resolved: facebookresearch#172

The FBNet V2 consists of several basic blocks of pattern conv(2d) + BatchNorm + ReLU. These operators can be fused into one operator, greatly speeding up the computation in GPU.

This diff / PR adds option to build an FBNet V2 backbone in fused manner, i.e. it looks up recursively the blocks and sub-blocks for the known patterns, and fuse them using `mobile_cv.arch.utils.fuse_utils`

Differential Revision: D47395999

fbshipit-source-id: d7b67d4d48e39ec887aa87393455528f6e519600
antoine-tran added a commit to antoine-tran/d2go that referenced this pull request Jul 13, 2023
Summary:
Pull Request resolved: facebookresearch#592

X-link: facebookresearch/mobile-vision#172

The FBNet V2 consists of several basic blocks of pattern conv(2d) + BatchNorm + ReLU. These operators can be fused into one operator, greatly speeding up the computation in GPU.

This diff / PR adds option to build an FBNet V2 backbone in fused manner, i.e. it looks up recursively the blocks and sub-blocks for the known patterns, and fuse them using `mobile_cv.arch.utils.fuse_utils`

Differential Revision: D47395999

fbshipit-source-id: b3da40ade8b36238dc9920c59a638357d861f283
@facebook-github-bot
Copy link
Contributor

This pull request was exported from Phabricator. Differential Revision: D47395999

antoine-tran added a commit to antoine-tran/d2go that referenced this pull request Jul 13, 2023
Summary:
Pull Request resolved: facebookresearch#592

X-link: facebookresearch/mobile-vision#172

The FBNet V2 consists of several basic blocks of pattern conv(2d) + BatchNorm + ReLU. These operators can be fused into one operator, greatly speeding up the computation in GPU.

This diff / PR adds option to build an FBNet V2 backbone in fused manner, i.e. it looks up recursively the blocks and sub-blocks for the known patterns, and fuse them using `mobile_cv.arch.utils.fuse_utils`

Differential Revision: D47395999

fbshipit-source-id: 406b08af4a238de95d126e2761d6b6896b9c8397
Summary:
X-link: facebookresearch/d2go#592

Pull Request resolved: facebookresearch#172

The FBNet V2 consists of several basic blocks of pattern conv(2d) + BatchNorm + ReLU. These operators can be fused into one operator, greatly speeding up the computation in GPU.

This diff / PR adds option to build an FBNet V2 backbone in fused manner, i.e. it looks up recursively the blocks and sub-blocks for the known patterns, and fuse them using `mobile_cv.arch.utils.fuse_utils`

Differential Revision: D47395999

fbshipit-source-id: 0f52567d7448ef30151422937a8af5efd6c44645
@facebook-github-bot
Copy link
Contributor

This pull request was exported from Phabricator. Differential Revision: D47395999

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

CLA Signed Do not delete this pull request or issue due to inactivity. fb-exported

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants