How It Works
Shard
One piece of a model, dataset or database that has been split across several machines so no single machine has to hold or process all of it. Large AI models are often sharded across many GPUs during training and inference.
Origin · no single documented coiner
A long-standing database and distributed-computing term for a horizontal partition of data. The word's computing use is often traced to the 1990s online game Ultima Online, whose separate server copies were called shards; this attribution is widely repeated but not formally documented.
No single primary source link for this term.