Skip to content

Conversation

littletomatodonkey
Copy link
Contributor

PR types

New features

PR changes

Models

Description

Support layout model to remove visual backbone for speed-up. Specifially,

  • LayoutLMv2(SER 84.5%, 31.46ms) -> LayoutLMv2-wo-backbone(SER 82.1%, 27.80ms)
  • LayoutXLM(SER 89.5%, 19.49ms) -> LayoutXLM-wo-backbone(SER 90.5%, 15.49ms)

Note The time cost is tested on V100 GPU for trt+fp16

@littletomatodonkey littletomatodonkey changed the title [WIP]add support for wo visual backbone add support for wo visual backbone Aug 3, 2022
Copy link
Contributor

@LiuChiachi LiuChiachi left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM

Comment on lines +300 to +301
"layoutlmv2-wo-backbone-base-uncased":
"https://bj.bcebos.com/paddlenlp/models/transformers/layoutlmv2/layoutlmv2-wo-backbone-base-uncased/model_state.pdparams",
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

这里的权重上传了嘛

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

模型均已上传~

Comment on lines +299 to +300
"layoutxlm-wo-backbone-base-uncased":
"https://bj.bcebos.com/paddlenlp/models/transformers/layoutxlm_wo_backbone_base/model_state.pdparams",
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

同上

@LiuChiachi LiuChiachi merged commit 54eda93 into PaddlePaddle:develop Aug 10, 2022
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants