TensorRT Optimizations

[CasanovaSan]

Will SDWebUI going to have native TensorRT support? (what i means is, will sdwebui install all of the necessary files for tensorrt and for the models be automatically converted for tensorrt and things like that) (I think it would be a good step for performance enhancement

[viperwasp]

I have no idea how this works. I updated my drivers. I think there was tensor update before and only models created using an updated method were compatible. I have no idea how to answer your question I only wanted to add these details. I too would love to see any improvements that will make this run faster. I would love to see answers to your questions.

[SplitMilky]

https://civitai.com/articles/2599/100percent-speed-boost-in-automatic1111-for-rtx-gpus-optimizing-checkpoints-with-tensorrt-extension

Guide to using the TensorTR.

Though I would appreciate some support from automatic1111 as I get a few errors with this guide. it worked great for 1.5sd but when I try with juggernautXL issues developed.

[mbloflin]

I used this much more thorough article on installing TensorRT and troubleshooting problems. I had a bunch of errors and it would crash A1111, but this guide got me going.

https://civitai.com/articles/2610/the-tensorrt-automatic1111-article-install-set-up-troubleshoot-technical-details

I would like a little more info on actually running it and all the options. It seems as if for me with 3080 TI, I have a lot more VRAM available than when using standard Models. Would love to see this running on SDXL (which is theoretically possible) but can't find much about it. Hopefully in a few weeks we will see some updates to make all this less cutting-edge.

[camoody1]

I know that the Dev channel of Automatic1111 has support for SDXL TensorRT, but when will the master channel be updated to support it? I'm desperate for a faster experience with SDXL.