Tag
25 articles
Learn how Baidu's Unlimited OCR achieves efficient processing of dozens of document pages in a single pass by mimicking human memory and forgetting mechanisms.
Learn to implement key Transformer architecture components including attention mechanisms and multi-head attention using PyTorch, replicating the technology behind OpenAI's successful AI systems.
Learn how to work with transformer architectures using Python and Hugging Face's transformers library. This tutorial teaches you to load models, process text, and understand the core components of modern AI systems like Google's Gemini.
Learn to implement a simplified version of MiniMax's Sparse Attention mechanism that reduces computational complexity in attention operations while maintaining performance.
Learn how xFormers helps make AI models faster and more memory-efficient by optimizing how they process text data.
This article explains NVIDIA's Nemotron 3.5 ASR, a 600M-parameter streaming speech recognition model that processes 40 languages in real-time using cache-aware optimization techniques.
Explains the advanced AI technologies behind Apple's upcoming Siri improvements in iOS 27, including transformer architectures, neural network enhancements, and privacy-preserving techniques.
Learn to set up and run inference with NVIDIA's Nemotron 3 Ultra, a 550B parameter hybrid Mamba-Transformer model designed for long-running AI agents with extended context windows.
Learn how fused kernels and automatic mixed precision (AMP) techniques, such as those in NVIDIA Apex and PyTorch's torch.amp, can dramatically accelerate transformer training by optimizing computational efficiency and reducing memory overhead.
Learn to implement and evaluate a hybrid MoE-diffusion model that demonstrates the performance benefits of converting autoregressive LLMs into diffusion models for improved inference speed.
Meta and Stanford researchers introduce the Fast Byte Latent Transformer, reducing inference memory bandwidth by over 50% without subword tokenization.
Learn how to set up and use FlashKDA, an open-source high-performance implementation of Kimi Delta Attention from Moonshot AI, for accelerating attention computation in large language models.
