วีดีโอ

Wohoo. That looks awesome!❤

a great talk!

Thanks for the forecast! Could you help me with something unrelated: I have a SafePal wallet with USDT, and I have the seed phrase. (alarm fetch churn bridge exercise tape speak race clerk couch crater letter). What's the best way to send them to Binance?

Great audience, loved the energy! Here is the abstract: "Large NLP models are dazzling us with their awesome performance. However, their size can make them cumbersome to handle, making experimentation more challenging and increases costs. Let’s use knowledge distillation to leverage the power of a high capacity teacher model to help a smaller student model to learn faster and performer better than it could on its own. The student model can even use a different network architecture than the teacher, one that fits the downstream task perfectly. We’ll see step-by-step how knowledge distillation works and how we can mix and match architectural features that are needed to optimize the model performance. What impact does for example the type and length of the input has on choosing the right tokenizer and neural network architecture?"

very good talk, thanks for sharing!

bravo czech people !

first

well done

At 10:14, why do H and W are much smaller than the original resolution of the document? Is there a some sort of compression or are the all zero embeddings eliminated? Edit: Ok I think I got it. 15-20 secs later you said that the image can be downsampled with 1/a where a is the size the smallest char

👏

Fantastic talk!