May 2026 monthly summary focused on documentation-driven delivery across PyTorch and ExecuTorch. Key features delivered: - PyTorch API Documentation Improvements: implemented extensive docs updates across MTIA memory, core PyTorch APIs, distributed checkpoint APIs, and Tensor Parallel APIs. Specifics include: - MTIA memory: Documented torch.mtia.memory.reset_peak_memory_stats and updated coverage/rendering directives; migrated relevant docs to MyST Markdown and ensured accurate rendering in MTIA memory docs. - Core/public APIs: Documented six public torch.utils.data APIs, added Sphinx autodoc directives, and fixed inline markup for clean rendering. - Distributed checkpoint: Documented seven public APIs with autodoc directives and adjusted docs conf to remove those entries from coverage_ignore_functions; verified rendering in docs/build/html/distributed.checkpoint.html. - Tensor Parallel: Documented three public APIs with autodoc directives and updated relevant docs page accordingly. - Documentation improvements included converting notes/out.rst to MyST Markdown (notes/out.md) and ensuring compatibility with Sphinx/Lint checks. - ExecuTorch Documentation Enhancement: Updated guidance for exporting custom LLMs to ExecuTorch, focusing on compatibility, KV cache ownership, and model boundaries. This is a docs-only PR that preserves runtime behavior and APIs. Major bugs fixed: - No code defects fixed in this period; several documentation quality issues resolved, including markup fixes and corrected examples to ensure accurate HTML rendering and searchability. - Resolved a few doc-rendering inconsistencies (RST vs MyST) and aligned cross-references for multi-repo docs. Overall impact and accomplishments: - Significantly improved developer onboarding and API discoverability through thorough documentation across MTIA memory, data utilities, distributed APIs, and tensor parallel features. - Strengthened documentation quality, consistency, and build reliability (validated with make coverage and make html). - Enhanced cross-project collaboration by documenting public APIs across PyTorch and ExecuTorch and updating shared docs conventions. Technologies/skills demonstrated: - Sphinx autodoc directives, MyST Markdown conversion, and autodoc directive usage across multiple modules. - Documentation build validation (make html, make coverage), doc lint checks, and conf.py adjustments for coverage controls. - Cross-repo documentation governance and maintenance, with emphasis on clarity, coverage, and visibility of public APIs.