Skip to content

[Blog]: multimodal blog - #470

Open
capri-xiyue wants to merge 4 commits into
llm-d:mainfrom
capri-xiyue:capri-xiyue/mm-blog
Open

[Blog]: multimodal blog#470
capri-xiyue wants to merge 4 commits into
llm-d:mainfrom
capri-xiyue:capri-xiyue/mm-blog

Conversation

@capri-xiyue

Copy link
Copy Markdown
Contributor

What does this PR do?

Why is this change needed?

How was this tested?

  • Tests added/updated (npm test)
  • Site builds successfully (npm run build:all)
  • Check links after buildling (npm run check-links)
  • Manual testing performed (npm run serve)

Checklist

  • Commits are signed off (git commit -s) per DCO
  • Code follows project contributing guidelines
  • Tests pass locally (npm test)
  • Site builds without errors (npm run build:all)
  • No broken links after building full site (npm run check-links)
  • Documentation updated (if applicable)

Related Issues

Signed-off-by: Xiyue Yu <xiyue@google.com>
@netlify

netlify Bot commented Aug 11, 2026

Copy link
Copy Markdown

Deploy Preview for llm-d ready!

Name Link
🔨 Latest commit 0721816
🔍 Latest deploy log https://app.netlify.com/projects/llm-d/deploys/6a7cb32fb438fd0008717f13
😎 Deploy Preview https://deploy-preview-470--llm-d.netlify.app
📱 Preview on mobile
Toggle QR Code...

QR Code

Use your smartphone camera to open QR code link.

To edit notification comments on pull requests, go to your Netlify project configuration.

Signed-off-by: Xiyue Yu <xiyue@google.com>
@capri-xiyue capri-xiyue changed the title WIP: multimodal blog docs: multimodal blog Aug 12, 2026
@capri-xiyue

Copy link
Copy Markdown
Contributor Author

assign @ahg-g

@capri-xiyue

Copy link
Copy Markdown
Contributor Author

assign @roytman


Reach for disaggregation when the encoder is genuinely the bottleneck: a high multimodal-to-text ratio, several large assets per request, or a ViT heavy enough that running it inline visibly stalls decode.

And be honest about the crossover. SGLang's report is blunt on this, and it applies equally here: for image-light workloads the network latency of shipping embeddings can outweigh the time saved by offloading, **raising** TTFT relative to collocation while dedicated encoder GPUs sit idle. Disaggregation is most efficient exactly when visual processing is the primary bottleneck. If it is not, you are paying for a tier and a network hop to solve a problem you do not have.

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

which SGLang report are you referring here? Maybe add a link to it for clarity.

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

added

Signed-off-by: Xiyue Yu <xiyue@google.com>
Signed-off-by: Xiyue Yu <xiyue@google.com>
@davidgs davidgs changed the title docs: multimodal blog [Blog]: multimodal blog Aug 13, 2026
@davidgs davidgs added the Blog Post This PR is a blog post label Aug 13, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

Blog Post This PR is a blog post

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants