I AM from india

This is really helpful it is surprising that very few people have checked this video out. Shows the market is more focused on cloud based LLMs rather than private LLMS that guarantee data security. This video is will do well in the coming future when the tides shift.

❤

I have trained a new transformer model (dinov2) for image classification on own data and locally saved the checkpoint. Now I want to convert this dinov2 model into onnx. Since the saved checkpoint is a directory containing several files (no a single .pth file), how to convert it to onnx?

is this leveraging UE Gbuffer at all?

Amazing

Hi! Thank you! Maybe you know same angular example ?

Thank you for doing this ! It would be such a treat if you could re-visit this with some examples in 5.3. perhaps something using depth maps from a webcam ? Cheers and hope you return for one more round. b

I'm impressed by how hilarious you guys are. Keep it up!

why would I use ONNX? why does it exist?

Hello, do you know how to convert a SafeTensor XL model to an ONNX XL model?

Live phi demo played again xd

The simplest most straightforward explanation for sampling, logits, top_k, top_p and temperature I saw. It takes really deep understanding to explain something so complex so simply. Thanks!

Great demo, thank you!

Is there an example of using phi3 vision model on Android?

OK. Now show us how its done!

NEED HELP! is there any way to convert a .tflite sparse model to onnx? I've tried several times but sparse models just don't convert.

Why was the pronunciation of Chat"GPT" censored? Around 1:39

Great content and exciting to see! Is there an example using Maui?

Nice. How to convert it to onnx using cuda?

I found a repo on hugging face that has an onnx version of falcon 7b (microsoft/falcon-7B-onnx) but it doesn't have the genai_config.json file in it. Do you know how to get that file for falcon?

thanks for this! just started using that style transfer model and got it working in the rust `ort` crate

How to inferenceSession on webgpu ?

Next level stuff here - for those who build AI models definitely worth converting them to ONNX and use the runtime features so edge devices interact with ONNX runtime w/o lag ... total game changer!

THIS IS AMAZING

can you make an example adout super resolution in unreal, please~~

I got exception : "Exception thrown: read access violation. Ort::GetApi(...) returned nullptr." on "FeatureMatcher->InitOrtEnv(cfg);"

sorry to say, but audio quality is pretty bad

Can someone help me convert a pt to onnx v8 ? :( not getting it done

Excellent demo. But the approach is still unclear in terms of final rendering. How do we see the predictions or output from the ML visualized in the UR? This is the first I am seeing on youtube though...

Thanks for the help

Thanks

How would it be useful to us if we cannot see the codes at all. And the github example file is also not available.

IS it valid for AMD GPUs also? like swap torch to torch-directml?

Can we add the links shown in the video, To the description of tis KZbin video. It'll be easy for those who'd want to follow along and do more research.

It will be good if you can display the examples you have shown in full screen

Thank You! All this is fantastic, however, please can you, Team MS, work on the stability of the OnnxRuntime and supporting Libs! Please, because I have struck so many issues, would be such a good resource if the stability was improved. I have logged some github issues already. This is such a cool tech! Thank you for all your very good videos!

Hi, thank you for the video! When I run this code I get a confidence of 0.701 instead of 0.921 like you do. Is there a reason for the discrepancy? I am using OnnxRuntime 1.16.3, CUDA 11.8, and cuDNN 8.7.0.

Requirements.txt has local references. Even when removing them it fails.

Can I build it for mobile game.Like gpu of mobile can be used by model.

what in the fck..why is it so complex -_-

Whether ONNX Runtime supports bare metal embedded platforms such as STM32

This is not "in the browser". This is still node.js. That's server technology.

I am trying to load the ssd mobilenet v2 tensorflow object detection onnx converted model into unreal engine 5. unreal is crashing. i need to know whether unreal 5 supports ssd mobilenet v2 object detection, If it works would you please instruct me to load the onnx model into unreal 5

there is a bug- InvalidArgument: [ONNXRuntimeError] : 2 : INVALID_ARGUMENT : Unexpected input data type. Actual: (tensor(int32)) , expected: (tensor(int64))

Any update on how to set it up with UE 5.3? There is no NNI plugin anymore.

This is amazing thanks for the quick intro to onnx runtime in Azure funapp.

You can't well train but try it, because you don't understand things and want to train it!!!

On line no 85 how did you get variable inputNames?

Great to come across this but would be really nice to have some longer tutorials, especially with explicit steps and demo projects to follow along with.