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.

Training MoE VLMs in Megatron Bridge v1.0.0

Use this resource to choose between FSDP and 3D parallelism to train a MoE VLM

New

  • 497 downloads
  • 138 likes
  • 219 saves

Free

Install with the CLI

In a terminal, from the folder you want:

npx --yes palatekit@latest install moe-vision-language-training-guide

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 recommends starting with FSDP for a first reliable run, then moving to 3D parallelism only once the PP layout is stable and there is time for tuning. It names the key knobs and configuration families. You get a training strategy.

Mostly picked by

Updated November 2, 2026

Card for the “Training MoE VLMs in Megatron Bridge” skill

Details

Works in
Install guides
Generates with
Your agent's model
Licence
Apache-2.0

Similar resources

View all

What you get

  • SKILL.md: comparison and decision guide
  • card.yaml: structured card
  • evals/evals.json: evaluation cases

Good fit for

  • You train a MoE VLM for the first time
  • Your multimodal model has a pipeline layout that is hard to split

Not for

    Tags

    • VLM
    • MoE
    • FSDP
    • multimodal

    About

    You say

    “Should I use FSDP or 3D parallelism for my MoE VLM on GB200?”

    Card for the “Training MoE VLMs in Megatron Bridge” skill

    Skill published by nvidia under the Apache-2.0 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 choose between FSDP and 3D parallelism to train a MoE VLM
    3. Review and adjust the result

    Content preview

    Files

    • BENCHMARK.md4.3 KB
    • card.yaml3 KB
    • evals/evals.json1.7 KB
    • LICENSE11.3 KB
    • NOTICE-PALATEKIT.md0.5 KB
    • skill-card.md3.9 KB
    • SKILL.md5.2 KB
    • skill.oms.sig4.7 KB

    Origin and licence

    Listed unmodified

    Author
    nvidia
    Source
    github.com/nvidia/skills/tree/ca116d60beafe1b4459f0ab790b7b450912c56b0/skills/nemo-mbridge-perf-moe-vlm-training
    Licence
    Apache-2.0

    Notice

    « nemo-mbridge-perf-moe-vlm-training » (nvidia, Apache-2.0) 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.