docs/guides/troubleshooting.md: broken example — Enable gradient checkpointing with training.gradient_checkpointing: true (DOC-D2B05-008) #199

Wave 1 · theme D1-10: Documentation-only fixes · highest severity High · 1 finding(s) · effort docs fix

Summary

Copying the snippet breaks the config.

Primary finding

DOC-D2B05-008 — docs/guides/troubleshooting.md: broken example — Enable gradient checkpointing with training.gradient_checkpointing: true

severity High · already tracked: KNOWN-OPEN C31-DOC-SCHEMA-PROSE — docs/roadmap/phase-16-trust-surface-hardening.md:785-786

Where. docs/guides/troubleshooting.md:159-164 (Turkish mirror: docs/guides/troubleshooting-tr.md:166-170) · type: broken example

The document says. Enable gradient checkpointing with training.gradient_checkpointing: true

What the code does. No such field exists: extra_forbidden validation error, exit 1 (verified). Gradient checkpointing is already enabled automatically whenever CUDA is available.

Code references. forgelm/trainer.py:514

User impact. Copying the snippet breaks the config.

Proposed correction. "Gradient checkpointing is enabled automatically on CUDA; no YAML key is needed (or available)."

Turkish mirror. the Turkish page has the same error; fix both in the same PR

Acceptance criteria

• Each statement is re-checked against the code at fix time.
• Documentation listed above is corrected in English and in its Turkish mirror in the same PR.
• Every finding in this issue is resolved or explicitly closed with a reason.

_{Found by the October 2026 ForgeLM code and documentation review against commit cab2947. Code links point at that commit; line numbers may have moved since. Finding IDs (CR-…, DOC-…, TECH-…) are stable references for this backlog.}