Machine Learning / AI

What are Tensor Cores and what operations do they accelerate?

Tell us about LongRAG and its idea of long search units.

Why is the VGG architecture not used today?

What is expert routing and what load balancing methods are there among experts?

Middle — Senior

HeadHunterМагнитМТС AI

Is it necessary to perform sampling close to reality or with a balanced class distribution when training YOLO for rare object detection?

What is SARSA and how does it differ from Q-learning?

Which ClickHouse engines are used and known?

Why did the problem of vanishing gradients occur earlier?

What are imitation learning and behavior cloning?

Middle — Senior

AstonМагнитПИК

What is the reason for the improvement in quality during self-model distillation?

Who has a larger receptive field: two consecutive 3x3 convolutions or one 5x5 convolution?

What is sharding in ClickHouse?

How to distill one model into another?

What is data distribution in ClickHouse?

What is a residual block?

Middle

AvitoМагнит

What are the metrics for evaluating dialogues (BLEU, perplexity, human eval)?