Daniel M. García-Ocaña Hernández

Sure, here is my code: ``` ``` and is embedded in a wordpress page. I took this example from [Gradio docs](https://www.gradio.app/guides/dynamic-apps-with-render-decorator#:~:text=Let%27s%20take%20a%20look%20at%20one%20last%20example%20that%20uses%20everything%20we%20learned.%20Below%20is%20an%20audio%20mixer.%20Provide%20multiple%20audio%20tracks%20and%20mix%20them%20together.).

Hi @stdKonjac! Similar question in #87

@bfshi also interested in knowing about 😄

Hi @cokeshao, why you saying this? In the figure it is clear that it improves LLM backbone performance in both TTFT and throughput: 

Oh, I see, my screenshot was regarding video input. So maybe the figure was obtained by only considering temporal token compression, or is a errata and for image input they...