Skip to content

Implement embedding bag convertor #2345

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
Closed
Christina-Young-NVIDIA opened this issue Sep 25, 2023 · 5 comments · Fixed by #2395 or #2677
Closed

Implement embedding bag convertor #2345

Christina-Young-NVIDIA opened this issue Sep 25, 2023 · 5 comments · Fixed by #2395 or #2677
Assignees
@zewenli98
Labels
component: converters Issues re: Specific op converters feature request New feature or request

Comments

@Christina-Young-NVIDIA
Copy link
Collaborator

No description provided.

Closed
@narendasan narendasan added component: converters Issues re: Specific op converters feature request New feature or request labels Sep 25, 2023
@zewenli98 zewenli98 mentioned this issue Oct 12, 2023
Merged
7 tasks
@narendasan narendasan reopened this Jan 2, 2024
@narendasan
Copy link
Collaborator

https://docs.nvidia.com/deeplearning/tensorrt/developer-guide/index.html#work-with-loops
https://github.com/pytorch/pytorch/blob/main/torch/nn/functional.py#L2368-L2391
pytorch/pytorch#25469

Helpful links

@narendasan
Copy link
Collaborator

  1. 1D input with DD-offsets (ITensor) -> TensorRT logic directly (ILoop)
  2. 2D input -> flattening, generating the offsets on the fly, then running 1D and reshape
  3. Undefined 2D behavior, unclear what happens in this case in PyT

@narendasan
Copy link
Collaborator

pytorch/pytorch#25469

@zewenli98
Copy link
Collaborator

TensorRT team just replied: "There's a known bug 4411383 where a network with DDS but not DS requires an optimization profile. For that bug, the easy way to avoid it in 9.2 was to enable profile sharing (PROFILE_SHARING_0806)."

@narendasan
Copy link
Collaborator

https://github.com/pytorch/pytorch/blob/8ca8729321a8c858c6bc33318ce2b80b8a5c900e/torch/onnx/symbolic_opset11.py#L1309

This was referenced Mar 8, 2024
Closed
Merged
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees

@zewenli98 zewenli98

Labels
component: converters Issues re: Specific op converters feature request New feature or request
Projects
None yet
Milestone
No milestone
3 participants
@narendasan @zewenli98 @Christina-Young-NVIDIA