Skillget

strix-halo-skills

skill
Guvenlik Denetimi
Uyari
Health Uyari
  • License — License: MIT
  • Description — Repository has a description
  • Active repo — Last push 0 days ago
  • Low visibility — Only 7 GitHub stars
Code Gecti
  • Code scan — Scanned 3 files during light audit, no dangerous patterns found
Permissions Gecti
  • Permissions — No dangerous permissions requested

Bu listing icin henuz AI raporu yok.

SUMMARY

Claude Code skill for AMD Strix Halo (Ryzen AI MAX+ 395) ML setup. Handles PyTorch installation (official wheels don't work with gfx1151), GTT memory config, and environment setup. Enables 30B parameter models.

README.md

Strix Halo PyTorch Skill

A Claude Code skill for setting up and validating PyTorch ROCm environments on
AMD Strix Halo (gfx1151) Linux systems.

The skill adds value beyond generic setup guidance by encoding the choices and
failure modes that are specific to this hardware:

  • AMD-supported and TheRock multi-arch installation tracks
  • real kernel checks rather than GPU-enumeration-only verification
  • forced PyTorch flash-attention validation with AMD's AOTriton opt-in
  • bounded FP16, BF16, MIOpen, backward, and torch.compile probes
  • correct unified-memory accounting and conservative GTT guidance
  • symptom-specific recovery without blanket HSA/ROCm overrides

Use

Copy .claude/skills/strix-halo-setup/ into a Claude Code project, then ask
Claude to set up or diagnose Strix Halo PyTorch. The skill starts with:

./.claude/skills/strix-halo-setup/scripts/verify_system.sh

For direct use in this repository, follow
SKILL.md.

Contents

.claude/skills/strix-halo-setup/
|-- SKILL.md
|-- docs/
|   |-- GTT_MEMORY_FIX.md
|   |-- INSTALLATION.md
|   |-- PERFORMANCE_FEATURES.md
|   `-- TROUBLESHOOTING.md
`-- scripts/
    |-- configure_gtt.sh
    |-- verify_pytorch.py
    `-- verify_system.sh

Scope

The skill covers Linux host checks, PyTorch installation, core capability
validation, and unified-memory setup. Model recipes, framework integrations,
Windows setup, and performance benchmarks are intentionally out of scope.

The TheRock track's current-release claims were checked on a Ryzen AI MAX+ 395
system with 64 GiB RAM. The AMD-supported track follows AMD's published matrix
and exact wheel set but was not reinstalled during that run. See
CHANGELOG.md for the tested software and results.

Sources

License

MIT License. Copyright (c) 2025 Ian Barber.

Yorumlar (0)

Sonuc bulunamadi