Every premium and members-only resource for $9.99/month, as a no-commitment subscription. See the plans

PalateKit

Sign in
Menu

Request a resource

Describe the resource you are missing in one sentence. We will email you when it is published.

0 / 280

A page, a video or an image online.

Unlock PalateKit

Unlock everything

Plan

Paying means accepting our terms of sale.

Secure payment by Stripe. Receipts and invoices sent by email.

Welcome back

Sign in to install, download or save this resource. A free account is enough for free resources.

Pretraining an LLM at scale with torchtitan v1.0.0

Use this resource to pretrain an LLM across several GPUs with 4D parallelism

New

  • 508 downloads
  • 183 likes
  • 103 saves

Free

Install with the CLI

In a terminal, from the folder you want:

npx --yes palatekit@latest install torchtitan-distributed-pretraining

With no option, the skill goes to ~/.agents/skills/; add --agent claude-code (or codex, cursor…) to target an agent.

First time: connect the CLI with

npx --yes palatekit@latest login

The agent guides you through installing torchtitan, configuring a Llama-style pretraining run on one or several nodes, then enabling Float8 and distributed checkpoints. It also covers multi-node training with SLURM.

Mostly picked by

Updated November 2, 2026

Card for the “Pretraining an LLM at scale with torchtitan” skill

Details

Works in
Install guides
Generates with
Your agent's model
Licence
MIT

Similar resources

View all

What you get

  • SKILL.md: installation, TOML configuration, single-node and multi-node SLURM launch
  • references/fsdp.md, float8.md, checkpoint.md and custom-models.md: FSDP2, Float8, checkpoints and custom models

Good fit for

  • You pretrain a model from scratch on several GPUs
  • You want Float8 on H100 GPUs

Not for

    Tags

    • pretraining
    • distributed
    • pytorch
    • fsdp

    About

    You say

    “Launch a Llama 3.1 8B pretraining on one 8-GPU node with torchtitan.”

    Card for the “Pretraining an LLM at scale with torchtitan” skill

    Skill published by Orchestra Research under the MIT license, delivered in its original version.

    Show more

    How to use it

    1. Install the skill in your agent's skills folder
    2. Ask your agent to pretrain an LLM across several GPUs with 4D parallelism
    3. Review and adjust the result

    Content preview

    Files

    • LICENSE1.1 KB
    • NOTICE-PALATEKIT.md0.5 KB
    • references/checkpoint.md4.3 KB
    • references/custom-models.md7.5 KB
    • references/float8.md4.1 KB
    • references/fsdp.md3.9 KB
    • SKILL.md9.1 KB

    Origin and licence

    Listed unmodified

    Author
    Orchestra Research
    Source
    github.com/orchestra-research/ai-research-skills/tree/773a52944ba4747a18bd4ae9ade53fff041adcbc/01-model-architecture/torchtitan
    Licence
    MIT

    Notice

    « distributed-llm-pretraining-torchtitan » (Orchestra Research, MIT) livré dans sa version d'origine : fiche en français et en anglais par PalateKit.

    User reviews

    –

    No ratings yet

    Loading reviews…

    Delete your review?

    Your rating and comment will be removed from the page.