spark-training-gotchas SKILL.md skillA

spark-training-gotchas from wshobson/agents is an agent skill: a SKILL.md folder an AI agent loads when a task calls for it.

Install it, pinned
mdr add wshobson/agents/spark-training-gotchas@git:20260714.baa5bd7

Pins git:20260714.baa5bd7 in mdr.lock. Pin the fingerprint instead.

Audit grade
A, 17 of 17 checks passed
Kind
SKILL.md skill
Repository
wshobson/agents
Path
plugins/dgx-spark-ops/skills/spark-training-gotchas/SKILL.md
Versions recorded
1
Latest version
git:20260714.baa5bd7, committed Jul 14, 2026
Size
7,981 bytes
Repository stars
42,320
License
MIT

Preflight and diagnose the ten known failure modes for ML training on NVIDIA DGX Spark. Use when a training run on DGX Spark fails to start, OOMs below the 128GB limit, slows down mid-run, or before any multi-hour training job on GB10.

Collected from public GitHub. Every version is saved exactly as it was and never changes. Install it at an exact version with the mdr tool, and every file is checked before it reaches your agent. The main file is checked against the SHA-256 fingerprint recorded here, and the others against the fingerprints GitHub keeps for its source commit. The automatic checks below grade the latest version, and the same checks always give the same file the same grade.

What the file says

# Spark Training Gotchas

DGX Spark's GB10 chip (Grace Blackwell, SM121, 128GB unified
memory, aarch64) has ten recurring failure modes across
launch, memory, thermals, bandwidth, and precision. Each is
named G1–G10 so it can be checked by number — the numbering
is load-bearing for tooling that runs these checks. Read this
before a long run, not after hour six.

## When to Use This Skill

- A training run fails to start, with an import error or a
  segfault that doesn't point at the real cause.
- A run OOMs while `nvidia-smi` still shows headroom.
- Throughput degrades partway through a run that started fine.
- Before any multi-hour or multi-epoch job on GB10.
- Wiring two Sparks together, before picking a parallelism
  strategy.
- Choosing between FP8 and NVFP4 for a Spark-hosted run.

## Common Issues Quick Reference

| # | Symptom | Fix |
|---|---|---|
| G1 | undefined symbol / segfault | cu130 wheel or container |
| G2 | flash-attn wrong backend used | skip pip build; monkeypatch on NGC |
| G3 | OOM despite headroom | drop page cache |
| G4 | throughput drop / reboot | expect ~100W sustained cap |
| G5 | memory-bound step slow | budget 180–192 GB/s |
…

Read the whole file at its exact version.

How to install

The exact contents, by fingerprint
mdr add wshobson/agents/spark-training-gotchas@sha256:4ba2e6510c325877

The line at the top of this page pins the label git:20260714.baa5bd7, which is what most people want. This one pins the sha256 fingerprint of the contents instead, so it survives a label being moved. Either way the exact version is written to mdr.lock, and mdr install downloads and checks those contents again, so it installs them exactly or stops with an error.

Badge

mdr badge

[![mdr](https://markdownregistry.com/badge/art_jvt64k46z57itqtw.svg)](https://markdownregistry.com/a/art_jvt64k46z57itqtw)

4 badge image fetches in 30 days. What the badge shows.

Versions

The registry has recorded 1 version of this file, committed on Jul 14, 2026.

versioncommittedcommitsizeaudit
git:20260714.baa5bd7 latest2026-07-14 baa5bd7 7,981 BA view

Audit of the latest version

A  17 of 17 checks passed. Fixed automatic checks with no AI involved, so the same checks give the same answer every run.
  • pass: Frontmatter block present
  • pass: Frontmatter declares a name
  • pass: Frontmatter declares a description
  • pass: Size between 200 bytes and 200 KB (7981 bytes)
  • pass: No zero-width or bidi control characters
  • pass: No instruction hidden inside an HTML comment
  • pass: No link to an exfiltration or paste host
  • pass: No credential-shaped string
  • pass: No instruction to send local credentials anywhere
  • pass: No text hidden with inline styles
  • pass: No prompt-injection phrasing
  • pass: No curl or wget piped into a shell
  • pass: No recursive delete of root, home or parent
  • pass: No instruction to read or print local credentials
  • pass: No base64 blob over 200 characters
  • pass: No link to a raw IP address
  • pass: No script tag

Source

GitHub

wshobson/agents · 42,320 stars · license MIT · pushed 2026-11-08 · branch main

API

GET https://markdownregistry.com/api/v1/artifacts/art_jvt64k46z57itqtw
GET https://markdownregistry.com/api/v1/resolve?ref=wshobson/agents/spark-training-gotchas
GET https://markdownregistry.com/api/v1/blob/4ba2e6510c3258776e01fe9419c427a81d5af5383ad64f3307957c47f3a5a6b8

Your AI assistant does the legwork. You hear about the deals worth your word. Hand yours the standing instructions at modelranch.com and it joins the network that reads files like this one.

Guides: the SKILL.md format and its required fields · what to check before you install a skill · how to pin a skill to an exact version · where to find agent skills · keeping a team on the same skills · which agent file does what.

More from wshobson/agents

000-project-conventions Cursor rule
wshobson/agents · .cursor/rules/000-project-conventions.mdc · claude-agents marketplace project conventions
git:20260707.0bacec2 · audit A · 42,320 stars
010-python-tooling Cursor rule
wshobson/agents · .cursor/rules/010-python-tooling.mdc · Python tooling and dependency conventions for plugin-eval and tools/
git:20260529.608c384 · audit A · 42,320 stars
020-agent-skill-authoring Cursor rule
wshobson/agents · .cursor/rules/020-agent-skill-authoring.mdc · Authoring agent, skill, and command markdown for cross-harness portability
git:20260913.1533e77 · audit A · 42,320 stars
AGENTS.md AGENTS.md file
wshobson/agents · AGENTS.md
git:20261015.7257c27 · audit A · 42,320 stars
CLAUDE.md CLAUDE.md file
wshobson/agents · CLAUDE.md
git:20260529.608c384 · audit B · 42,320 stars
screen-reader-testing SKILL.md skill
wshobson/agents · plugins/accessibility-compliance/skills/screen-reader-testing/SKILL.md · Test web applications with screen readers including VoiceOver, NVDA, and JAWS. Use when validating screen reader…
git:20261015.13703c9 · audit A · 42,320 stars
wcag-audit-patterns SKILL.md skill
wshobson/agents · plugins/accessibility-compliance/skills/wcag-audit-patterns/SKILL.md · Conduct WCAG 2.2 accessibility audits with automated testing, manual verification, and remediation guidance. Use when…
git:20260522.be57c0b · audit A · 42,320 stars
multi-reviewer-patterns SKILL.md skill
wshobson/agents · plugins/agent-teams/skills/multi-reviewer-patterns/SKILL.md · Coordinate parallel code reviews across multiple quality dimensions with finding deduplication, severity calibration…
v1.0.2 · audit A · 42,320 stars
parallel-debugging SKILL.md skill
wshobson/agents · plugins/agent-teams/skills/parallel-debugging/SKILL.md · Debug complex issues using competing hypotheses with parallel investigation, evidence collection, and root cause…
v1.0.2 · audit A · 42,320 stars
parallel-feature-development SKILL.md skill
wshobson/agents · plugins/agent-teams/skills/parallel-feature-development/SKILL.md · Coordinate parallel feature development with file ownership strategies, conflict avoidance rules, and integration…
v1.0.2 · audit A · 42,320 stars
task-coordination-strategies SKILL.md skill
wshobson/agents · plugins/agent-teams/skills/task-coordination-strategies/SKILL.md · Decompose complex tasks, design dependency graphs, and coordinate multi-agent work with proper task descriptions and…
v1.0.2 · audit A · 42,320 stars
team-communication-protocols SKILL.md skill
wshobson/agents · plugins/agent-teams/skills/team-communication-protocols/SKILL.md · Structured messaging protocols for agent team communication including message type selection, plan approval, shutdown…
v1.0.2 · audit A · 42,320 stars

Every file in wshobson/agents

Browse by kind, by grade A, or by owner.