English
← Back to AI Technology

Training and post-training

Distributed Training

Techniques that partition data, parameters, activations, or computation across devices and nodes. They make large-model training possible but introduce communication, synchronization, and fault-tolerance tradeoffs.

Reference period
2012–present
Last reviewed

Key terms

  • data parallelism
  • tensor parallelism
  • pipeline parallelism
  • sharding

Connected across the project

Primary sources

  1. ImageNet Classification with Deep Convolutional Neural Networks
  2. Efficient Large-Scale Language Model Training on GPU Clusters Using Megatron-LM
  3. PyTorch Fully Sharded Data Parallel

This is a curated technical map, not a claim of comprehensive coverage.