Domains
Forms
Mechanism
Breaks down how the skill works and what it produces, so you can quickly judge whether it fits your scenario and value.
When a training run on NVIDIA DGX Spark (GB10) fails to start, OOMs below the 128GB limit, slows down mid-run, or before any multi-hour training job on GB10.
assets/preflight.sh automates checks for G1, G3, G4, G7, G9 producing PASS/FAIL/WARN/SKIP/INFO lines.references/gotcha-checks.md holds detailed manual check commands per gotcha.A fast preflight diagnosis and remediation guide that prevents wasted multi-hour runs by catching CUDA ABI, OOM, thermal, and parallelism misconfigurations before launch.
Use Cases
A ML engineer starts a LoRA fine-tune on DGX Spark but the process OOMs while nvidia-smi shows free memory. They invoke this skill, run assets/preflight.sh which flags G3 (UMA OOM below 128GB). Following the fix, they drop the page cache with sync; echo 3 > /proc/sys/vm/drop_caches and the run proceeds. This saves hours of failed attempts and protects the unified memory pool.
Before a team kicks off a 10-hour multi-epoch job on GB10, they run the skill's preflight to catch G1 (CUDA 13 ABI mismatch) and G9 (bare pip drift). The check reveals wheels link libcudart.so.12; they switch to an NGC container per spark-environment-setup. This prevents a mid-run segfault and ensures the expensive job completes without environment breakage.
Skill Relationships
Dependency relationships read as "the upper tier points to the lower tier." The current Skill sits in the middle tier — above are Skills that depend on it, below are Skills it depends on.
Tier 1 · These Skills Use Me
Tier 2 · Current Skill
Tier 3 · I Use These Skills
Browse this skill's relationships within its skillset. Click a node to switch the side panel; use the search box to jump to any skill.
Skill File