Revision History
Tensor Parallelism · 7 revisions
Sizes are character counts of the article source. The signed number is the change from the previous revision.
Recent edit summaries
Detailed summaries recorded by editors. Generic maintenance summaries are omitted here; every recorded revision remains below. Dates describe the edit, not necessarily the event it covers.
- Correction: Corrected the Megatron-Turing NLG configuration (no 24-way data parallelism or 6,720 GPUs in the paper), removed an invalid TP=72 inference claim with unsourced latency figures, and fixed the wrong author list for the Flash Communication reference (plus 8 minor fixes)
Version 7 · Nov 2, 2026, 04:01 AM
- Correction: TorchTitan TP degree was "1 to 16"; its paper says TP is usually capped at 8 within a node; removed unsupported "selective TP"; linked /wiki/torchtitan
Version 6 · Oct 16, 2026, 11:06 PM
- Improve internal links (2026-10-13)
Version 5 · Oct 13, 2026, 10:18 AM
- Added 6 contextual internal links
Version 4 · Jul 23, 2026, 03:28 PM