What models are you using, or hoping to use, with TensorRT?

[jaybdub]

Hi All,

We're interested in hearing how you're using, or hoping to use torch2trt. This helps us prioritize features, or suggest workarounds if they exist. Please feel free to share here with

Using: <model name>, <model URL>
Notes: ...

or

Hoping to Use: <model name>, <model URL>
Notes: ...

If you find a model in this thread that matches you're use case, simply +1 that comment.

Look forward to hearing from you!

Best,
John

[Boatsure]

Hoping to Use: CenterNet, https://openaccess.thecvf.com/content_ICCV_2019/html/Duan_CenterNet_Keypoint_Triplets_for_Object_Detection_ICCV_2019_paper.html

[jaybdub]

Thanks @surefyyq ! Hope to take a look soon. Please feel to reach out if you have other suggestions as well.

[nanmi]

You can refer to this repo: https://github.com/nanmi/CenterNet-trt

[YoushaaMurhij]

I hope to use CenterPoint (CVPR paper)
The model contains multiple modules and a multiple inputs/outputs(heads) .. including dictionary data structures.

[sunziq322]

I want to deploy a customed yolov5m.pt on jetson NX.
Jetpack 4.5, cuda 10.2, cudnn8.0, tensorrt 7.3.xx
Is that possible to convert it to .engine?

[Fan9]

Bert !!!!!

[Deep-learning999]

Hoping to Use: YOLOP SLOWFAST SLOWONLY ST-GCN YOLOV5 DEEPSORT

[imoneoi]

Hope to use: DeepLabv3+

[FreedomLiX]

Write codes in C++ API

[maronuu]

Hoping to Use: YOLOR, paper

[rodja]

@maronuu has made some progress with YoloR and described it in #565. Now it seems the torch.nn.parameter.Parameter converter is missing...

[tianyu06030020]

hoping for: torch.nn.ReflectionPad2d(), which is a padding implemention from pytorch. thx

[Roulbac]

Hoping to use Spatial Transformer Networks : link

[prabhd33ps]

Hoping to use DSFD Official Link

[xunleiwu]

Hoping to Use: DeTR, https://github.com/facebookresearch/detr
Thank you very much.
We are looking for the conversion plugin of Multi-Head Attention used in DeTR.

[peria1]

Hoping to Use: Detectron2, https://github.com/facebookresearch/detectron2

We have had trouble with other routes to production, so we are taking a look at TensorRT.

[qzsrh]

Hoping to Use: yolact
https://github.com/KevinJia1212/yolact_cityscapes_550

[iNeil77]

Hoping to use DeiT variant of VIT
https://arxiv.org/abs/2012.12877

[Deep-learning999]

Hoping to Use: YOLOP SLOWFAST SLOWONLY ST-GCN YOLOV5 DEEPSORT

[Deep-learning999]

https://github.com/yysijie/st-gcn

[oliver0922]

mobilenetv2 Please :)
Link: https://pytorch.org/vision/master/_modules/torchvision/models/mobilenetv2.html

[SJavad]

I also have to convert ssd mobile net to tensorrt model for inferencing on jetson tx1

[Hong-Bo]

StyleGAN2-pytorch Danke!
Link: https://github.com/rosinality/stylegan2-pytorch

[FreedomLiX]

Take it easy

…

---Original--- From: ***@***.***&gt; Date: Wed, Nov 9, 2022 15:52 PM To: ***@***.***&gt;; Cc: ***@***.******@***.***&gt;; Subject: Re: [NVIDIA-AI-IOT/torch2trt] What models are you using, or hoping to use, with TensorRT? (Discussion #531) YOLOv7-pose — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: ***@***.***&gt;

[FreedomLiX]

code link

…

---Original--- From: ***@***.***&gt; Date: Wed, Nov 9, 2022 15:52 PM To: ***@***.***&gt;; Cc: ***@***.******@***.***&gt;; Subject: Re: [NVIDIA-AI-IOT/torch2trt] What models are you using, or hoping to use, with TensorRT? (Discussion #531) YOLOv7-pose — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: ***@***.***&gt;

[TiesenX]

Hoping to Use: Detic
https://github.com/facebookresearch/Detic

[FreedomLiX]

please info the net details!

…

---Original--- From: "Le Minh ***@***.***&gt; Date: Mon, Feb 6, 2023 12:38 PM To: ***@***.***&gt;; Cc: ***@***.******@***.***&gt;; Subject: Re: [NVIDIA-AI-IOT/torch2trt] What models are you using, or hoping touse, with TensorRT? (Discussion #531) Hoping to Use: Detic https://github.com/facebookresearch/Detic — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: ***@***.***&gt;

[blue-q]

Hoping to Use: FIDTM
https://github.com/dk-liang/FIDTM

[foemre]

Hoping to use: MiDaS
https://github.com/isl-org/MiDaS

[ziqin-h]

Hoping to use: Bot-Sort
https://github.com/NirAharon/BoT-SORT

[RemanentQuill76]

Hoping to use: Yolov8 detect
https://hub.ultralytics.com/models/Wif16r2vflrJ1rc30ca2

[fabricecarles]

Hoping to Use MiDaS https://pytorch.org/hub/intelisl_midas_v2/
Hoping to Use RAFT https://pytorch.org/vision/main/models/raft.html

[lofyol]

yolov8-pose

[kais-bedioui]

@lofyol Hi, you might want to keep an eye on this repo: https://github.com/marcoslucianops/DeepStream-Yolo-Pose The author mentioned lately (on Linkedin at least) that he will start working on Yolo-Pose Good Luck

…

On Mon, 11 Sep 2023 at 09:57, lofyol ***@***.***> wrote: yolov8-pose — Reply to this email directly, view it on GitHub <#531 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AI7SGIERQQT6WHIJE7TTGNLXZ275RANCNFSM4ZQ45EUQ> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.*** com>

[soerenbrockmann]

How to convert a model that has not supported operations? Is there any example for it? Documentation is quite hard to understand...

[mukul74]

Hoping to Use: Retinexformer

[ahmet-f-gumustas]

Hoping to Use : DDRNet

[yachty66]

Hoping to Use : Flux Fill Dev