hi@aiweekly.co.in

Telegram CEO accuses Reliance of ‘sabotaging access’ for users outside India, says ‘this may be part of competitive war’

Telegram CEO Pavel Durov criticised Reliance for allegedly using BGP hijacking to block Telegram for users outside India. He linked it to a broader competitive conflict amid Telegram’s ban ahead of NEET UG exam. 

Telegram CEO accuses Reliance of ‘sabotaging access’ for users outside India, says ‘this may be part of competitive war’ Read More »

How to Build Memory-Efficient Transformers with xFormers Using Packed Sequences, GQA, ALiBi, SwiGLU, and Causal Attention

In this tutorial, we implement xFormers: a practical toolkit for building fast, memory-efficient Transformer models on GPUs. We begin by validating memory-efficient attention against a standard attention implementation, then compare their speed and memory consumption across different sequence lengths. We then examine causal masking, packed variable-length sequences, grouped-query attention, and custom ALiBi positional biases. Finally,

How to Build Memory-Efficient Transformers with xFormers Using Packed Sequences, GQA, ALiBi, SwiGLU, and Causal Attention Read More »