Skip to content
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

Backport compatibility with TensorRT version 10 from yolov8 #12984

Merged
merged 1 commit into from
May 5, 2024
Merged

Backport compatibility with TensorRT version 10 from yolov8 #12984

merged 1 commit into from
May 5, 2024

Conversation

n1mmy
Copy link
Contributor

@n1mmy n1mmy commented May 5, 2024 •
edited by github-actions bot
Loading

Based on the is_trt10 code in yolov8.

🛠️ PR Summary

Made with ❤️ by Ultralytics Actions

🌟 Summary

Introducing compatibility enhancements for TensorRT 10 in the YOLOv5 export process.

📊 Key Changes

  • Added checks to determine if the TensorRT version is 10 or above.
  • Modified the way memory workspace size is set for version 10.
  • Adjusted engine building process to support changes in TensorRT 10.
  • Updated model bind and context setting methods to accommodate both pre-version 10 and version 10 TensorRT logic.

🎯 Purpose & Impact

  • 🚀 Purpose: To ensure YOLOv5's export script remains compatible with the latest advancements in NVIDIA's TensorRT optimization toolkit, specifically version 10.
  • 💡 Impact: Users will benefit from continued support and enhancements in model optimization and deployment. This makes the transition to TensorRT 10 seamless for developers, potentially leading to improved inference performance and efficiency on NVIDIA GPUs.

@n1mmy
Add compatibility with TensorRT version 10.
653cbae 
Based on the is_trt10 code in yolov8.
@github-actions GitHub Actions
Copy link
Contributor

github-actions bot commented May 5, 2024 •
edited
Loading

CLA Assistant Lite bot All Contributors have signed the CLA. ✅

@n1mmy
Copy link
Contributor Author

n1mmy commented May 5, 2024

I have read the CLA Document and I sign the CLA

@glenn-jocher
Copy link
Member

@n1mmy thank you for your contributions :)

@glenn-jocher glenn-jocher merged commit 920c721 into ultralytics:master May 5, 2024
9 checks passed
thebusted added a commit to thebusted/yolov5 that referenced this pull request May 7, 2024
@thebusted
Merge pull request #1 from ultralytics/master
0fdde04 
Backport compatibility with TensorRT version 10 from yolov8 (ultralytics#12984)
kevinconka added a commit to SEA-AI/yolov5 that referenced this pull request May 28, 2024
@kevinconka @franziska-denk
@GilSeamas @glenn-jocher @snyk-bot @UltralyticsAssistant @dependabot @n1mmy @inisis
Horizon detection using YOLOv5's backbone (#2)
458b802 
* Modified iou values and extended hyps

* Add pre-augmentation step for 16bit thermal imgs

* training with 16to8bit augmentation works

* Modifications to 16-to-8 bit augmentations

* minor rename

* minor changes due to usage

* Grouped training scripts in folder

* added horizon utils

* Added dataloaders

* rename to avoid confusion w/ yolov5/utils folder

* add pitch and yaw targets

* HorizonDataset working w albumentations transform

* Training loop working (needs refactoring)

* Create README.md

Signed-off-by: Kevin Serrano <35779409+kevinconka@users.noreply.github.com>

* refactored models.py

* Added TODOs to README

* added train.py and transforms.py to cleanup nb

* Checked some TODOs from horizon/README.md

* updated tests nb

* Save val/train loss plots as png images

* Added loss weigths for horizon training

* Added more TODOs

* Added weather augmentations to rgb

* tensorrt engine working (WIP)

* added patch for tensorRT v7

* Update README.md

Signed-off-by: Kevin Serrano <35779409+kevinconka@users.noreply.github.com>

* added small change for wandb run_id

* Update models.py

Signed-off-by: Kevin Serrano <35779409+kevinconka@users.noreply.github.com>

* Added batch support for yolotrt

* folder refactoring

* more refactoring... added "hybrids" folder

* Update yolotrt.py

Signed-off-by: Kevin Serrano <35779409+kevinconka@users.noreply.github.com>

* working nb for ahoy-IR.engine

* AHOYv5 and DANv5 refactored

* typing compatibility with python3.8

* get images from url and use softmax for horizon postprocessing in dan_tests

* made output_modes compatible with batched inference for ahoy

* tweakes plus extended export function for AHOY and DAN models

* Add benchmarking script for AHOY and DAN models

* fixed misspelling

* new preprocessing function

* faster preprocessing

* fast preprocess

* Add close_mosaic option to train.py and dataloaders.py

* Update hyp.sea-ai-IR.yaml

* Update hyp.sea-ai.yaml

* Update close-mosaic argument description

* increase kmeans iter for autoanchor

* added class mapping to dan

* refactored dan cls_maps

* Update Discord and Contributing Guide URLs (ultralytics#12847)

* Update Discord and Contributing Guide URLs

* Update __init__.py

Signed-off-by: Glenn Jocher <glenn.jocher@ultralytics.com>

---------

Signed-off-by: Glenn Jocher <glenn.jocher@ultralytics.com>

* refactored trt inference code after adding it to Core-Backend

* fixed bug in horizon output decoding when lr padding was added during inference

* added small comment

* add export hooks to get fp32 to fp16 done in GPU

* register hooks before dry runs when exporting hybrids

* added preprocessing operations to the model via hooks before trt export

* added hook to export process

* make dummy input int8 for trt export

* added softmax to postprocessing hook to add it to the model

* remove softmax from theta, offset postprocessing

* [Snyk] Security upgrade pillow from 9.5.0 to 10.3.0 (ultralytics#12868)

* fix: requirements.txt to reduce vulnerabilities


The following vulnerabilities are fixed by pinning transitive dependencies:
- https://snyk.io/vuln/SNYK-PYTHON-PILLOW-6514866

* Update requirements.txt

Signed-off-by: Glenn Jocher <glenn.jocher@ultralytics.com>

---------

Signed-off-by: Glenn Jocher <glenn.jocher@ultralytics.com>
Co-authored-by: snyk-bot <snyk-bot@snyk.io>

* Create cla.yml (ultralytics#12899)

Signed-off-by: Glenn Jocher <glenn.jocher@ultralytics.com>

* Sort imports with Ruff and iSort (ultralytics#12915)

* Sort imports with Ruff and iSort

Signed-off-by: Glenn Jocher <glenn.jocher@ultralytics.com>

* Auto-format by https://ultralytics.com/actions

---------

Signed-off-by: Glenn Jocher <glenn.jocher@ultralytics.com>
Co-authored-by: UltralyticsAssistant <web@ultralytics.com>

* Create merge-main-into-prs.yml (ultralytics#12918)

* Create merge-main-into-prs.yml

Signed-off-by: Glenn Jocher <glenn.jocher@ultralytics.com>

* Update merge-main-into-prs.yml

Signed-off-by: Glenn Jocher <glenn.jocher@ultralytics.com>

---------

Signed-off-by: Glenn Jocher <glenn.jocher@ultralytics.com>

* Update to `ultralytics>=8.1.47` (ultralytics#12919)

Signed-off-by: Glenn Jocher <glenn.jocher@ultralytics.com>

* Update merge-main-into-prs.yml (ultralytics#12920)

Signed-off-by: Glenn Jocher <glenn.jocher@ultralytics.com>

* Update merge-main-into-prs.yml (ultralytics#12921)

Signed-off-by: Glenn Jocher <glenn.jocher@ultralytics.com>

* Update merge-main-into-prs.yml (ultralytics#12922)

Signed-off-by: Glenn Jocher <glenn.jocher@ultralytics.com>

* Bump gunicorn from 19.10.0 to 22.0.0 in /utils/google_app_engine (ultralytics#12929)

* Bump gunicorn from 19.10.0 to 22.0.0 in /utils/google_app_engine

Bumps [gunicorn](https://github.com/benoitc/gunicorn) from 19.10.0 to 22.0.0.
- [Release notes](https://github.com/benoitc/gunicorn/releases)
- [Commits](benoitc/gunicorn@19.10.0...22.0.0)

---
updated-dependencies:
- dependency-name: gunicorn
  dependency-type: direct:production
...

Signed-off-by: dependabot[bot] <support@github.com>

* [Snyk] Security upgrade gunicorn from 19.10.0 to 22.0.0 (ultralytics#12938)

fix: utils/google_app_engine/additional_requirements.txt to reduce vulnerabilities


The following vulnerabilities are fixed by pinning transitive dependencies:
- https://snyk.io/vuln/SNYK-PYTHON-GUNICORN-6615672

Co-authored-by: snyk-bot <snyk-bot@snyk.io>

---------

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
Co-authored-by: Glenn Jocher <glenn.jocher@ultralytics.com>
Co-authored-by: snyk-bot <snyk-bot@snyk.io>

* Refactor model variable names in export.py and export_hybrid.py

* Bump slackapi/slack-github-action from 1.25.0 to 1.26.0 in /.github/workflows (ultralytics#12948)

* Bump slackapi/slack-github-action in /.github/workflows

Bumps [slackapi/slack-github-action](https://github.com/slackapi/slack-github-action) from 1.25.0 to 1.26.0.
- [Release notes](https://github.com/slackapi/slack-github-action/releases)
- [Commits](slackapi/slack-github-action@v1.25.0...v1.26.0)

---
updated-dependencies:
- dependency-name: slackapi/slack-github-action
  dependency-type: direct:production
  update-type: version-update:semver-minor
...

Signed-off-by: dependabot[bot] <support@github.com>

* Update ci-testing.yml

Signed-off-by: Glenn Jocher <glenn.jocher@ultralytics.com>

---------

Signed-off-by: dependabot[bot] <support@github.com>
Signed-off-by: Glenn Jocher <glenn.jocher@ultralytics.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
Co-authored-by: Glenn Jocher <glenn.jocher@ultralytics.com>

* ResizeIfNeeded augmentation and plot MSEs during training

* Update num_workers in dataloaders.py

* Add Turkish and Vietnamese Docs (ultralytics#12972)

* Add Turkish and Vietnamese Docs

Signed-off-by: Glenn Jocher <glenn.jocher@ultralytics.com>

* Update README.md

Signed-off-by: Glenn Jocher <glenn.jocher@ultralytics.com>

* Update pyproject.toml

Signed-off-by: Glenn Jocher <glenn.jocher@ultralytics.com>

---------

Signed-off-by: Glenn Jocher <glenn.jocher@ultralytics.com>

* Refactor horizon.py and custom.py for better readability and maintainability.

* Add best_mse.pt checkpoint saving during training

* Fix data type conversion issue in custom.py

* Fix data type conversion issue in train.py

* wandb integration working

* cleaned nb

* scale pitch for gt and preds when processing 320x256 imgs

* fix msu_sum and better cli logging

* fixed mask visualisation

* cast to int32 for cv2.line

* Fix formatting issue in cli logging

* Backport compatibility with TensorRT version 10 from yolov8 (ultralytics#12984)

Add compatibility with TensorRT version 10.

Based on the is_trt10 code in yolov8.

* Set `TORCH_CPP_LOG_LEVEL=ERROR` for reduced verbosity (ultralytics#12989)

* modify torch cpp log level to Error to avoid annoying print

* Auto-format by https://ultralytics.com/actions

---------

Co-authored-by: UltralyticsAssistant <web@ultralytics.com>

* Add `pip install --retries 3` to CI to resolve transients (ultralytics#13001)

* Add `pip install --retries 3` to CI to resolve transients

Signed-off-by: Glenn Jocher <glenn.jocher@ultralytics.com>

* Update ci-testing.yml

Signed-off-by: Glenn Jocher <glenn.jocher@ultralytics.com>

---------

Signed-off-by: Glenn Jocher <glenn.jocher@ultralytics.com>

* Revert CI `pip install` retries to default (ultralytics#13002)

Signed-off-by: Glenn Jocher <glenn.jocher@ultralytics.com>

* Centralize ENV variable definition in utils/general.py (ultralytics#13004)

* Centralize ENV variable definition in utils/general.py

Signed-off-by: Glenn Jocher <glenn.jocher@ultralytics.com>

* Update export.py

Signed-off-by: Glenn Jocher <glenn.jocher@ultralytics.com>

* Auto-format by https://ultralytics.com/actions

---------

Signed-off-by: Glenn Jocher <glenn.jocher@ultralytics.com>
Co-authored-by: UltralyticsAssistant <web@ultralytics.com>

* Bump contributor-assistant/github-action from 2.3.2 to 2.4.0 in /.github/workflows (ultralytics#13006)

Bump contributor-assistant/github-action in /.github/workflows

Bumps [contributor-assistant/github-action](https://github.com/contributor-assistant/github-action) from 2.3.2 to 2.4.0.
- [Release notes](https://github.com/contributor-assistant/github-action/releases)
- [Commits](contributor-assistant/github-action@v2.3.2...v2.4.0)

---
updated-dependencies:
- dependency-name: contributor-assistant/github-action
  dependency-type: direct:production
  update-type: version-update:semver-minor
...

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

* Reformat Markdown code blocks (ultralytics#13023)

* [Snyk] Security upgrade requests from 2.31.0 to 2.32.0 (ultralytics#13039)

fix: requirements.txt to reduce vulnerabilities


The following vulnerabilities are fixed by pinning transitive dependencies:
- https://snyk.io/vuln/SNYK-PYTHON-REQUESTS-6928867

Co-authored-by: snyk-bot <snyk-bot@snyk.io>

* Update Ultralytics YouTube URL (ultralytics#13046)

* Update Ultralytics YouTube URL

* Update README.md

Signed-off-by: Glenn Jocher <glenn.jocher@ultralytics.com>

* Auto-format by https://ultralytics.com/actions

---------

Signed-off-by: Glenn Jocher <glenn.jocher@ultralytics.com>
Co-authored-by: UltralyticsAssistant <web@ultralytics.com>

* minor notebook update

* Function docstrings added

* Auto-format by https://ultralytics.com/actions

---------

Signed-off-by: Kevin Serrano <35779409+kevinconka@users.noreply.github.com>
Signed-off-by: Glenn Jocher <glenn.jocher@ultralytics.com>
Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: franziska-denk <denk.m.franziska@gmail.com>
Co-authored-by: GilSimas <gil.simas@sea.ai>
Co-authored-by: Glenn Jocher <glenn.jocher@ultralytics.com>
Co-authored-by: snyk-bot <snyk-bot@snyk.io>
Co-authored-by: UltralyticsAssistant <web@ultralytics.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
Co-authored-by: Nick Martin <284356+n1mmy@users.noreply.github.com>
Co-authored-by: inisis <46103969+inisis@users.noreply.github.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers
No reviews
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.

None yet
2 participants
@n1mmy @glenn-jocher