Interview Practice Questions

Не знаю, но мы так просто как бы не называли. Поэтому сложно сказать, это является этими или нет?

А у вас получается можно чад GPT прикрутить к инфраструктуре или как?

Как бы такую проблему ты предложил решить?

Tell about Transformer architecture

Does attention computational complexity differ between inference and training in decoder?

What is key-value caching in transformers?

Why can't we just run ML application on Bare Metal?

Why do we need VLM (Vision-Language Model) if we have regular Transformers?

What is Flash Attention and how does it work?

What implementations of gradient boosting exist?

Can you tell how different machine learning algorithms differ from each other?

What validation strategies exist for models?

Can a model degrade over time?

How to properly split data into training and test sets?

What is Bagging?

Tell about Random Forest

How many times will variance decrease in ideal case when using ensemble?

How does RMSProp optimizer work?

What minimum amount of video memory will be needed for model training?

What regularization methods in neural networks do you remember?