User tests whether Jev can generate images
A user asks @typesafeai's Jev whether it can generate images and shares the sunflower image Jev produced.
Integrating Jev into MiniMax H3 to Accelerate Video Generation
The author integrated Jev into MiniMax H3, using Jev to sparsify Attention processing by determining importance per layer and selecting sparsity rates of 1%, 3%, 5%, and 10%, reducing video generation time on an RTX4070 from 6 min 7 sec to 3 min.
I integrated the popular Jev into MiniMax H3 to speed up video generation! Using Jev for sparsification of Attention processing. ・Jev determines importance per layer (4-step, targeting 49 layers) ・Jev selects sparsity rates of 1%, 3%, 5%, 10% On RTX4070, from 6 min 7 sec → 3
· 249 likesOpen on X
More like this
Trying Jev to Evaluate Video Production Quality
The author tries Jev to see if better eval scores can actually lead to better video production quality, hoping the numbers translate to pixels.
Building a Visual Reference Finder with Jev
One prompt fetches 100 images from Cosmos, NASA, and The Met for a visual reference finder used in creative work.
