VoltaML

Appearance

AITemplate Acceleration ​

Benefits ​

  • About 1.7x faster generation

Downsides ​

  • No LoRA support
  • No Textual Inversion support
  • Locked down resolution (if you use static AITemplate)

Acceleration ​

Make sure that you are one the correct page as can be seen in the image below.

AITemplate Acceleration

  1. Width and Height - these cannot be changed after the model is compiled
  2. Batch Size - this cannot be changed after the model is compiled as well - but Batch Count can - for this reason, I recommend setting it to 1
  3. CPU Threads - Number of CPU threads that will be used for compilation - MORE THREADS, MORE RAM NEEDED - if you do not have enough RAM, lower this number down, 6-8 Threads is recommended (about 7GB free RAM needed)
  4. Model - model that will be accelerated
  5. Accelerate Button - click this to start the acceleration process

WARNING

Due to extreme load on the CPU, connection between the browser and the server will be lost. If this happens, you need to click the reconnect button - acceleration should not be affected.

WE WOULD RECOMMEND RESTARTING VOLTA AFTER ACCELERATION IS DONE, AS WEBSOCKETS MIGHT BE TOTALLY BROKEN

How to use ​

  1. Switch the Model tab to AITemplate
  2. Click the Load button
  3. Generate an image - only txt2img, img2img and ControlNet are supported

AITemplate Acceleration