nightly-bisect/a: [Bug Fix] Fix CPUAdagrad updates with nonzero weight decay (#8781)
Fix CPUAdagrad to use the weight-decay-adjusted gradient in both the accumulator and parameter update, matching torch.optim.Adagrad . …
What moved
Fix CPUAdagrad to use the weight-decay-adjusted gradient in both the accumulator and parameter update, matching torch.optim.Adagrad . Previously, the update numerator incorrectly used the original gradient. Covers scalar and SIMD implementations, including the XPU copies. …
Why it matters
Filed as a public note from DeepSpeed, 2026-10-11. The desk tags this brief Hardware.
On the record
- Filed from the DeepSpeed official GitHub releases feed on 2026-10-11.
- Primary source host: github.com.
- Fix CPUAdagrad to use the weight-decay-adjusted gradient in both the accumulator and parameter update, matching torch.optim.Adagrad .
Primary source
github.com
github.com
Desk
Logged as brief 109
Logged as brief 109
Cite this brief
nightly-bisect/a: [Bug Fix] Fix CPUAdagrad updates with nonzero weight decay (#8781) (DeepSpeed, 2026-10-11) — https://labledgerdesk.pages.dev/b/2026/10/11/nightly-bisect-a-bug-fix-fix-cpuadagrad-updates-with-nonzero-weight/