1 comment

[ 10.1 ms ] story [ 52.6 ms ] thread
A website that allows you to run any model precompiled and optimized on Jetson devices including live runs and runtime benchmarks.

An example is running Cosmos Reason 3 Edge on an Orin Nano at 152 tokens/s https://embedl.com/news/embedl-and-space-cargo-unlimited-bri...

Some other models coming up including VLMs and VLAs like Pi and Flux action.

Is this interesting to open up for the community as a counterpart to HuggingFace where you get models optimized and built, ready to run?