Armory
Source
Browse
Skills

distributed-llm-pretraining-torchtitan

Provides PyTorch-native distributed LLM pretraining using torchtitan with 4D parallelism (FSDP2, TP, PP, CP). Use when pretraining Llama 3.1, DeepSeek V3, or custom models at scale from 8 to 512+ GPUs with Float8, torch.compile, and distributed checkpointing.

Score
Unranked
Evidence
No signals yet
Last commit
Not known
Listed

Install

armory install distributed-llm-pretraining-torchtitan --cli claude

writes the skill into.claude/skills/distributed-llm-pretraining-torchtitan/SKILL.mdListed as compatible

Configuration
# fetches the source and writes it to:
.claude/skills/distributed-llm-pretraining-torchtitan/SKILL.md

Needs the armory CLI · not on npm yet, build it from cli/ in the repository

What it is

Provides PyTorch-native distributed LLM pretraining using torchtitan with 4D parallelism (FSDP2, TP, PP, CP). Use when pretraining Llama 3.1, DeepSeek V3, or custom models at scale from 8 to 512+ GPUs with Float8, torch.compile, and distributed checkpointing.

When to use it

Provides PyTorch-native distributed LLM pretraining using torchtitan with 4D parallelism (FSDP2, TP, PP, CP). Use when pretraining Llama 3.1, DeepSeek V3, or custom models at scale from 8 to 512+ GPUs with Float8, torch.compile, and distributed checkpointing.

How to install / invoke

# Copy the skill into your .claude/skills/ directory
curl -sL https://raw.githubusercontent.com/davila7/claude-code-templates/main/cli-tool/components/skills/ai-research/model-architecture-torchtitan/SKILL.md -o .claude/skills/model-architecture-torchtitan/SKILL.md

Notes

Extracted from davila7/claude-code-templates, model-architecture-torchtitan category.