Categories
Deep-Learning Mechanisms
Distributed Training
GPU Systems and Performance Engineering
Inference and Serving
Transformer Architecture
Evaluation and Experimentation
Multimodal Learning
Tokenization and Representation
Reliability and Observability
Software Architecture
2026
Tensor-Parallel GQA Attention, One Head Group at a Time
Megatron Tensor Parallelism: Rebuilding a SwiGLU Feed-Forward Block Across Two Ranks
Pixels Are Not Positions: Testing 2D Coordinate Encodings Under Scale and Translation
A Restart Is Not a Resume: Testing PyTorch Checkpoints
