Skip to content

[rhaiis] add custom dspark spec-decoding preset - #180

Open
aas008 wants to merge 2 commits into
openshift-psap:mainfrom
aas008:rhaiis-deepseek-v4-spec-decoding
Open

[rhaiis] add custom dspark spec-decoding preset#180
aas008 wants to merge 2 commits into
openshift-psap:mainfrom
aas008:rhaiis-deepseek-v4-spec-decoding

Conversation

@aas008

@aas008 aas008 commented Aug 19, 2026

Copy link
Copy Markdown

Add four H200 Zeus comparison presets: baseline, native MTP, DSpark fixed-k, and DSpark with dynamic SD. Adaptive verification is omitted because DeepSeek-V4 varlen full graphs are SM100-only.

Add four H200 Zeus comparison presets: baseline, native MTP, DSpark
fixed-k, and DSpark with dynamic SD. Adaptive verification is omitted
because DeepSeek-V4 varlen full graphs are SM100-only.

Co-authored-by: Cursor <cursoragent@cursor.com>
@openshift-ci

openshift-ci Bot commented Aug 19, 2026

Copy link
Copy Markdown

[APPROVALNOTIFIER] This PR is NOT APPROVED

This pull-request has been approved by:
Once this PR has been reviewed and has the lgtm label, please assign tosokin for approval. For more information see the Code Review Process.

The full list of commands accepted by this bot can be found here.

Details Needs approval from an approver in each of these files:

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

@coderabbitai

coderabbitai Bot commented Aug 19, 2026

Copy link
Copy Markdown

Important

  • 🔍 Trigger review

This repository does not receive automatic reviews because it has fewer than 10 stars.

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: e38dec0d-bc59-4642-9a43-f94b5d729d7b


Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

Comment @coderabbitai help to get the list of available commands.

Run profile1-4 and enable the vLLM profiler so spec-decoding jobs
land on the staging dashboard with traces uploaded to S3.

Co-authored-by: Cursor <cursoragent@cursor.com>
@aas008

aas008 commented Aug 19, 2026

Copy link
Copy Markdown
Author

/test fournos rhaiis deepseek-v4-dspark profile1
/pipeline forge-full
/cluster zeus
/var rhaiis.engines.vllm.images.nvidia: vllm/vllm-openai:v0.27.0
/var tests.rhaiis.version: vLLM-0.27.0

@psap-forge-bot

Copy link
Copy Markdown

🔴 Execution of rhaiis deepseek-v4-dspark profile1 🔴

Execution Engine Configuration

forge:
  args:
  - deepseek-v4-dspark
  - profile1
  configOverrides:
    rhaiis.engines.vllm.images.nvidia: vllm/vllm-openai:v0.27.0
    tests.rhaiis.version: vLLM-0.27.0
  project: rhaiis

Artifact Links

Test Logs

00 Pre-Cleanup 1 second

01 Prepare 3 seconds

02 Preflight 1 second

03 Test 1 hour, 5 minutes, 38 seconds

04 Post-Cleanup 6 seconds

🔄 05 Export-Artifacts

Post-processing Status

  • parse: failed * No records found - parsing completed successfully but no test data was extracted

@psap-forge-bot

Copy link
Copy Markdown
🔴 Submission of rhaiis deepseek-v4-dspark profile1 failed after 1 hour, 9 minutes, 18 seconds 🔴

Error: FournosJobFailureError: FOURNOS Job 'forge-rhaiis-20260819-223718' failed: Tasks Completed: 6 (Failed: 1, Cancelled 0), Skipped: 0

/test fournos rhaiis deepseek-v4-dspark profile1
/var rhaiis.engines.vllm.images.nvidia: vllm/vllm-openai:v0.27.0
/var tests.rhaiis.version: vLLM-0.27.0
/pipeline forge-full
/cluster zeus

engine_args:
tensor-parallel-size: 8

deepseek-v4-pro-0813: &deepseek-v4-pro-0813

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

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

what do the &, * < symbols here mean?

Copy link
Copy Markdown
Author

Choose a reason for hiding this comment

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

YAML anchors and aliases. I learnt its a way to write the shared model block once and reuse it. I used them here to keep the file succinct; & names the block, * copies it, and <<: merges it so the MTP/DSpark variants only add speculative-config.

Copy link
Copy Markdown
Author

Choose a reason for hiding this comment

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

In current architecture of this models.yaml, ever config would need a new entry. This will bloat up very very fast.

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

This will bloat up very very fast.

if you look how the presets work, something like:

common_values:
  abc: true

specific_values:
  extends: [common_values]
  efg: true

you can build a mechanism to avoid the config file size explosion. That may be more flexible than YAML anchors

or alternatively, if applicable, some default values that apply to all the models

default:
  engine_args:
    tensor-parallel-size: 8

- profile2
- profile3
- profile4
rhaiis.profiler.enabled: true

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

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

do we need profiling for spec decoding?

Copy link
Copy Markdown
Author

Choose a reason for hiding this comment

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

This is a one time use preset @Harshith-umesh , should I add a tag to not merge it to main?

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

you can comment /hold in the main thread to prevent the merge and/or add Draft flag in the PR to convey that

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

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

yes if it's a one time preset then you can add a do not merge tag to the PR

Copy link
Copy Markdown
Author

Choose a reason for hiding this comment

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

I don't have permissions to add the label.

@aas008

aas008 commented Aug 20, 2026

Copy link
Copy Markdown
Author

I'm trying to understand why my failed. Is it because ready_timeout: 3600, and the download + loading the weigths could not be done in 1 hr? @kpouget @Harshith-umesh ?
It will be common for RHAIIS users to run new models from PRs, can we extend this timeout? Actually I see I can set it myself - rhaiis.deploy.ready_timeout

I think the bigger thing was rhaiis.deploy.image_pull_secrets: ["npalaska-image-pull"] was not added by default.

@Harshith-umesh

Copy link
Copy Markdown
Member

nope the timeout is not this issue, please see this file on mlflow: mlflow-artifacts:/workspaces/forge-rhaiis/264/da8ad9a31b4e49e0ac57957ff6e02a8b/artifacts/03__test/001__test_deepseek-v4-pro-0813-dspark_profile1_profile2_profile3_profile4/002__capture_isvc_state/artifacts/inferenceservice.pods.logs

@aas008

aas008 commented Aug 20, 2026

Copy link
Copy Markdown
Author

Ah thank you I could download and see the memory issue.
Any reason I could not preview it on the web UI?

@aas008

aas008 commented Aug 20, 2026

Copy link
Copy Markdown
Author

/test fournos rhaiis deepseek-v4-dspark
/pipeline forge-full
/cluster zeus
/exclusive true
/var rhaiis.engines.vllm.images.nvidia: vllm/vllm-openai:v0.27.0
/var tests.rhaiis.version: vLLM-0.27.0

@aas008

aas008 commented Aug 20, 2026

Copy link
Copy Markdown
Author

/test cancel

@aas008

aas008 commented Aug 20, 2026

Copy link
Copy Markdown
Author

/test cancel

@Harshith-umesh Can I cancel a test?
image

@aas008

aas008 commented Aug 20, 2026

Copy link
Copy Markdown
Author

/test fournos rhaiis deepseek-v4-dspark
/pipeline forge-full
/cluster zeus
/exclusive true
/var rhaiis.engines.vllm.images.nvidia: vllm/vllm-openai:v0.27.0
/var tests.rhaiis.version: vLLM-0.27.0

@psap-forge-bot

Copy link
Copy Markdown

🛑 Execution of rhaiis deepseek-v4-dspark 🛑
🛑 JOB ABORTED - spec.shutdown=Stop

Execution Engine Configuration

forge:
  args:
  - deepseek-v4-dspark
  configOverrides:
    rhaiis.engines.vllm.images.nvidia: vllm/vllm-openai:v0.27.0
    tests.rhaiis.version: vLLM-0.27.0
  project: rhaiis

Artifact Links

Test Logs

00 Pre-Cleanup 1 second

01 Prepare 3 seconds

02 Preflight 1 second

03 Test

04 Post-Cleanup 10 seconds

🔄 05 Export-Artifacts

@kpouget

kpouget commented Aug 20, 2026

Copy link
Copy Markdown
Contributor

/test cancel

fyi, in the github workflow (/test ...) you abort a test by pushing a commit to the PR. On purpose or because you forget you had a test running (talking from experience 😅 )

@psap-forge-bot

Copy link
Copy Markdown

🔴 Execution of rhaiis deepseek-v4-dspark 🔴

Execution Engine Configuration

forge:
  args:
  - deepseek-v4-dspark
  configOverrides:
    rhaiis.engines.vllm.images.nvidia: vllm/vllm-openai:v0.27.0
    tests.rhaiis.version: vLLM-0.27.0
  project: rhaiis

Artifact Links

Test Logs

00 Pre-Cleanup 1 second

01 Prepare 3 seconds

02 Preflight 1 second

03 Test 1 hour, 5 minutes, 32 seconds

04 Post-Cleanup 7 seconds

🔄 05 Export-Artifacts

Post-processing Status

  • parse: failed * No records found - parsing completed successfully but no test data was extracted

@psap-forge-bot

Copy link
Copy Markdown
🔴 Submission of rhaiis deepseek-v4-dspark failed after 1 hour, 9 minutes, 36 seconds 🔴

Error: FournosJobFailureError: FOURNOS Job 'forge-rhaiis-20260820-203017' failed: Tasks Completed: 6 (Failed: 1, Cancelled 0), Skipped: 0

/test fournos rhaiis deepseek-v4-dspark
/var rhaiis.engines.vllm.images.nvidia: vllm/vllm-openai:v0.27.0
/var tests.rhaiis.version: vLLM-0.27.0
/pipeline forge-full
/cluster zeus
/exclusive true

@openshift-ci

openshift-ci Bot commented Aug 20, 2026

Copy link
Copy Markdown

@aas008: The following test failed, say /retest to rerun all failed tests or /retest-required to rerun all mandatory failed tests:

Test name Commit Details Required Rerun command
ci/prow/fournos be9efff link true /test fournos

Full PR test history. Your PR dashboard.

Details

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes-sigs/prow repository. I understand the commands that are listed here.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants