xyz-ai-lab
- 4 skills
- 0 followers
- 1 day ago last updated
- ▌ Axrl Fix Linting · xyz-ai-labFix lint, formatting, pyright, and mypy failures in this repository by running scripts/run-precommit-check.sh, making the smallest non-semantic changes, and stopping if a fix would change logits or training semantics.
- ▌ Axrl Spike Replay · xyz-ai-lab bundleReproduce and analyze AXRL gradient spike snapshots, run mitigation experiments on a GPU node, and document proven spike-mitigation methods.
- ▌ Axrl Run Regression Tests · xyz-ai-labRun axrl validation end-to-end by using skills/axrl-fix-linting/SKILL.md first, then skills/axrl-run-gpu-task/SKILL.md to execute scripts/run-tests.sh on a remote GPU node, iterating on failures and benchmark table updates.
- ▌ Axrl Review Staged Changes · xyz-ai-labReview git-staged changes (a specific file if given, otherwise all staged files) for correctness bugs, coding-style and codebase-consistency issues, and suggestions for modularization, maintainability, efficiency, and readability. Report findings without applying fixes unless the user asks.