PyTorch v2.14.1 randomness notes disclaim complete reproducibility across releases (and, per the same passage, commits and platforms), and the notes no longer mention CUBLAS_WORKSPACE_CONFIG, consistent with the producer-reported removal of the requirement in PR...

PyTorch v2.14.1 randomness notes disclaim complete reproducibility across releases (and, per the same passage, commits and platforms), and the notes no longer mention CUBLAS_WORKSPACE_CONFIG, consistent with the producer-reported removal of the requirement in PR #161749. Curator qualifications: PR #161749 itself, the C++ check removal and function-local static read of the env var (F5 remainder, F6) not reinspected; the per-process static claim is an unexecuted code reading. use_deterministic_algorithms docstring counts, SDPA backend and TF32 defaults (F3, F7, F9) not reinspected. NVIDIA still documents the env var for multi-stream determinism (CA-B023): keep both statements. Docs/headers only; no model run. Limitations: Docs/headers only; no model run. Evidence status: statements located in the cited sources by a correlated desk review (Claude); not executed; not independently reviewed.

{
  "citations": [
    {
      "url": "https://github.com/pytorch/pytorch",
      "note": "PyTorch v2.14.1 docs/source/notes/randomness.md line 5; search CUBLAS_WORKSPACE"
    }
  ]
}