Categories

Transformers

Diagram of a transformer layer with separate activation perturbations and rewards for four tokens

Dust claims transformers can be pretrained without backpropagation

An abstract model graph passes through a glass optimization lens into high-speed inference lanes

Hugging Face and vLLM bring native-speed serving to Transformers model definitions

A transformer-shaped lens maps scattered data points into smooth density contours

Allen AI's DiScoFormer tests one transformer for density and score

A mixture-of-experts model is split across GPUs while a single import path feeds the training pipeline

NVIDIA NeMo AutoModel makes MoE fine-tuning a one-import upgrade

NVIDIA's Hugging Face article shows NeMo AutoModel wrapping expert parallelism and custom kernels behind the familiar Transformers loading path for MoE fine-tuning.

The AI Feed Desk

By The AI Feed Desk

A testing gauge compares a clean tool path with a longer tangled debugging path

Hugging Face measures whether tools are agent-friendly

Hugging Face's agent-focused benchmark tests whether software changes help coding agents finish tasks with fewer errors, tokens, and detours.

The AI Feed Desk

By The AI Feed Desk