Skip to content

How does TransformerEngine interact with TensorRT/Onnx? #212

New issue
New issue
Closed
Closed
How does TransformerEngine interact with TensorRT/Onnx?#212

Description

@tylerweitzman
tylerweitzman
opened on May 10, 2023

Hi,
We are often converting PyTorch models to other engines in order to optimize inference speed. Is it possible to convert a PyTorch model that is using TransformerEngine into TensorRT for faster inference? Or is it a choice either between nn.Linear with TensorRT right now?

Activity

  1. nzmora-nvidia commented on May 13, 2023

    @nzmora-nvidia
    nzmora-nvidia
    on May 13, 2023
    Contributor

    Hi @tylerweitzman,

    We are evaluating support for exporting TE models to ONNX and import from TensorRT. Support for FP8 GPT export will be rolled out first.
    Can you share what kind of models and precisions are you interested in?

  2. NVIDIA
    locked and limited conversation to collaborators on May 22, 2023
  3. ksivaman
    converted this issue into a discussion #238 on May 22, 2023
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions