NVIDIA / NVIDIA/TensorRT

TensorRT 10.5 Flux Dit BF16 precision

Open
#4,215 5 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Module:Accuracy Module:Demo triaged
Dominant language
C++
Stars
13.4k
Forks
2.4k
Avg merge
5d 3h
Merged PRs (30d)
2

Description

Description

When I used TensorRT 10.5 to infer Flux Dit on A800 using BF16 dataType, I found that there was a significant decrease in accuracy, while there was no significant decrease in accuracy when I used Pytorch BF16 to infer

Environment

TensorRT Version:

NVIDIA GPU: A800

NVIDIA Driver Version: 535.54.03

CUDA Version:12.2

CUDNN Version:

Operating System:

Python Version (if applicable):

Tensorflow Version (if applicable):

PyTorch Version (if applicable):

Baremetal or Container (if so, version):

Relevant Files

Model link:

Steps To Reproduce

Commands or scripts:

Have you tried the latest release?:

Can this model run on other frameworks? For example run ONNX model with ONNXRuntime (polygraphy run <model.onnx> --onnxrt):

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start by collecting the missing TensorRT version, model link, commands or scripts, and complete environment details. Reproduce Flux Dit BF16 inference on the A800 with TensorRT and PyTorch, then compare their accuracy; done requires a reproducible accuracy difference and enough information to investigate it.

Written by the indexing model from the issue text.

Assessment

Tech stack
pytorch
Domain
machine-learning, performance
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.