[rhaiis] Add missing Model Furnace models and new NVFP4/Qwen3.x/DeepSeek models by ssaketh-ch · Pull Request #173 · openshift-psap/forge

ssaketh-ch · 2026-08-17T20:05:41Z

Summary

Add 15 models from Model Furnace that were missing from Forge, with TP
sizes and vllm args matching gitlab.cee.redhat.com/psap/model-furnace.
Add 14 new models not yet in either system: NVFP4 quantizations,
Qwen3.6 and Qwen3.8 families, and DeepSeek-V4-Flash.
Fix two broken preset aliases in presets.yaml that pointed at undefined
model keys (phi-4 and llama-405b).

Motivation

Model Furnace and Forge had diverged on which models each tracks. This
closes the gap for the rhaiis project and adds recently released models
on top.

Testing

Models added from Model Furnace have been validated there. New models
(NVFP4, Qwen3.6, Qwen3.8, DeepSeek-V4-Flash) will need a benchmark run
to confirm the TP values and vllm args are correct.

Models added

From Model Furnace (not yet in Forge)

Key	HF Model ID	TP
`phi-4`	`microsoft/phi-4`	1
`phi-4-w4a16`	`RedHatAI/phi-4-quantized.w4a16`	1
`phi-4-w8a8`	`RedHatAI/phi-4-quantized.w8a8`	1
`phi-4-fp8`	`RedHatAI/phi-4-FP8-dynamic`	1
`llama-3-1-8b-fp8`	`RedHatAI/Meta-Llama-3.1-8B-Instruct-FP8-dynamic`	1
`llama-3-1-8b-w8a8`	`RedHatAI/Meta-Llama-3.1-8B-Instruct-quantized.w8a8`	1
`llama-3-1-8b-w4a16`	`RedHatAI/Meta-Llama-3.1-8B-Instruct-quantized.w4a16`	1
`llama-3-1-405b`	`meta-llama/Llama-3.1-405B-Instruct`	8
`llama-3-1-405b-fp8`	`RedHatAI/Meta-Llama-3.1-405B-Instruct-FP8-dynamic`	8
`llama-3-1-405b-w8a8`	`RedHatAI/Meta-Llama-3.1-405B-Instruct-quantized.w8a8`	8
`qwen3-5-35b-fp8`	`RedHatAI/Qwen3.5-35B-A3B-FP8-dynamic`	1
`qwen3-5-122b-fp8`	`RedHatAI/Qwen3.5-122B-A10B-FP8-dynamic`	2
`qwen3-5-397b-fp8`	`RedHatAI/Qwen3.5-397B-A17B-FP8-dynamic`	4
`qwen3-coder-next-fp8`	`RedHatAI/Qwen3-Coder-Next-FP8-dynamic`	1
`infomaniak-translategemma-12b`	`Infomaniak-AI/vllm-translategemma-12b-it`	1

New models (not in Model Furnace or Forge)

Key	HF Model ID	TP
`kimi-k2-5-nvfp4`	`nvidia/Kimi-K2.5-NVFP4`	4
`glm-5-2-nvfp4`	`RedHatAI/GLM-5.2-NVFP4-FP8`	4
`nemotron3super-120b-nvfp4`	`RedHatAI/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4`	1
`deepseek-v4-pro-nvfp4`	`RedHatAI/DeepSeek-V4-Pro-NVFP4-FP8`	8
`gemma-4-26b-a4b-nvfp4`	`RedHatAI/gemma-4-26B-A4B-it-NVFP4`	1
`gemma-4-31b-nvfp4`	`RedHatAI/gemma-4-31B-it-NVFP4`	1
`kimi-k3-nvfp4`	`RedHatAI/Kimi-K3-NVFP4`	4
`qwen3-6-35b-a3b`	`Qwen/Qwen3.6-35B-A3B`	1
`qwen3-6-35b-fp8`	`RedHatAI/Qwen3.6-35B-A3B-FP8`	1
`qwen3-6-35b-nvfp4`	`RedHatAI/Qwen3.6-35B-A3B-NVFP4`	1
`qwen3-8-27b`	`Qwen/Qwen3.8-27B`	2
`qwen3-8-27b-fp8`	`Qwen/Qwen3.8-27B-FP8`	1
`qwen3-8-2-4t-fp8`	`RedHatAI/Qwen3.8-2.4T-A95B-FP8`	8
`deepseek-v4-flash-nvfp4`	`RedHatAI/DeepSeek-V4-Flash-NVFP4-FP8`	4

Made with Cursor

… models Adds models that were in Model Furnace but not yet in Forge, and new models not yet in either. TP sizes and vllm args for the Model Furnace entries match the source at gitlab.cee.redhat.com/psap/model-furnace. The new models include NVFP4 quantizations, the Qwen3.6 and Qwen3.8 families, and DeepSeek-V4-Flash. Co-authored-by: Cursor <cursoragent@cursor.com>

coderabbitai · 2026-08-17T20:05:53Z

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: fd098f25-52e4-4a22-9462-51e521b3bcaf

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.}

ssaketh-ch · 2026-08-17T20:10:34Z

@Harshith-umesh, @aas008, @naveenmiriyaluredhat. Cant add anyone as a reviewer for some reason.

naveenmiriyaluredhat · 2026-08-17T21:12:53Z

@sacherukuri
Please add the following :

…veen's request Co-authored-by: Cursor <cursoragent@cursor.com>

aas008 · 2026-08-18T01:04:19Z

/test fournos rhaiis nvidia benchmark hera ci-quick
/pipeline forge-full
/exclusive false
/cluster hera

ssaketh-ch · 2026-08-18T13:32:10Z

/test fournos rhaiis nvidia benchmark hera ci-quick
/pipeline forge-full
/exclusive false
/cluster hera

psap-forge-bot · 2026-08-18T13:40:14Z

🟢 Execution of `rhaiis` nvidia benchmark hera ci-quick 🟢

Execution Engine Configuration

forge:
  args:
  - nvidia
  - benchmark
  - hera
  - ci-quick
  configOverrides: {}
  project: rhaiis

Artifact Links

Test Logs

✅ 00 Pre-Cleanup `1 second`

✅ 01 Prepare `2 seconds`

✅ 02 Preflight `1 second`

✅ 03 Test `5 minutes, 14 seconds`

✅ 04 Post-Cleanup `4 seconds`

🔄 05 Export-Artifacts

Post-processing Status ✅

✅ parse: success
✅ artifacts_to_kpis: success
- 📄 kpis.json
✅ kpis_to_mlflow: success
✅ kpis_to_csv: success
- 📊 dashboard.csv
⏭️ artifacts_to_ai_data: disabled

kpi.artifacts_to_ai_data disabled
⏭️ s3_import: disabled

s3_import disabled
⏭️ analyse_kpis: disabled

analyze disabled
⏭️ s3_export: disabled

s3_export disabled

psap-forge-bot · 2026-08-18T13:40:32Z

🟢 Submission of rhaiis nvidia benchmark hera ci-quick succeeded after 7 minutes, 48 seconds 🟢

/test fournos rhaiis nvidia benchmark hera ci-quick
/pipeline forge-full
/exclusive false
/cluster hera

Adds the three missing GuideLLM workload profiles from guidellm.yaml. Co-authored-by: Cursor <cursoragent@cursor.com>

ssaketh-ch · 2026-08-18T18:21:20Z

/test fournos rhaiis nvidia benchmark hera ci-quick
/pipeline forge-full
/exclusive false
/cluster hera

psap-forge-bot · 2026-08-18T18:29:35Z

🟢 Execution of `rhaiis` nvidia benchmark hera ci-quick 🟢

Execution Engine Configuration

forge:
  args:
  - nvidia
  - benchmark
  - hera
  - ci-quick
  configOverrides: {}
  project: rhaiis

Artifact Links

Test Logs

✅ 00 Pre-Cleanup `1 second`

✅ 01 Prepare `2 seconds`

✅ 02 Preflight `1 second`

✅ 03 Test `5 minutes, 15 seconds`

✅ 04 Post-Cleanup `4 seconds`

🔄 05 Export-Artifacts

Post-processing Status ✅

✅ parse: success
✅ artifacts_to_kpis: success
- 📄 kpis.json
✅ kpis_to_mlflow: success
✅ kpis_to_csv: success
- 📊 dashboard.csv
⏭️ artifacts_to_ai_data: disabled

kpi.artifacts_to_ai_data disabled
⏭️ s3_import: disabled

s3_import disabled
⏭️ analyse_kpis: disabled

analyze disabled
⏭️ s3_export: disabled

s3_export disabled

psap-forge-bot · 2026-08-18T18:29:53Z

🟢 Submission of rhaiis nvidia benchmark hera ci-quick succeeded after 7 minutes, 47 seconds 🟢

/test fournos rhaiis nvidia benchmark hera ci-quick
/pipeline forge-full
/exclusive false
/cluster hera

albertoperdomo2 · 2026-08-19T14:18:41Z

/lgtm

albertoperdomo2 · 2026-08-19T14:19:23Z

/lgtm
/approve

openshift-ci · 2026-08-19T14:19:29Z

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: albertoperdomo2

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

The pull request process is described here

Details

Needs approval from an approver in each of these files:

~~OWNERS~~ [albertoperdomo2]

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

feat: add gemma-4-26B-A4B, Muse-Glimmer-30B, and Inkling-Small per na…

cd089e7

…veen's request Co-authored-by: Cursor <cursoragent@cursor.com>

feat: add profile5 profile6 profile7 from Model Furnace

caadafd

Adds the three missing GuideLLM workload profiles from guidellm.yaml. Co-authored-by: Cursor <cursoragent@cursor.com>

Merge branch 'main' into feat/add-missing-and-nvfp4-models

ad32d5e

albertoperdomo2 closed this Aug 19, 2026

albertoperdomo2 reopened this Aug 19, 2026

openshift-ci Bot assigned albertoperdomo2 Aug 19, 2026

openshift-ci Bot added the lgtm Indicates that a PR is ready to be merged. label Aug 19, 2026

openshift-ci Bot added the approved Indicates a PR has been approved by an approver from all required OWNERS files. label Aug 19, 2026

openshift-merge-bot Bot merged commit fb2b054 into openshift-psap:main Aug 19, 2026
9 of 10 checks passed

Uh oh!

Conversation

ssaketh-ch commented Aug 17, 2026

Summary

Motivation

Testing

Models added

From Model Furnace (not yet in Forge)

New models (not in Model Furnace or Forge)

Uh oh!

coderabbitai Bot commented Aug 17, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

ssaketh-ch commented Aug 17, 2026

Uh oh!

naveenmiriyaluredhat commented Aug 17, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

aas008 commented Aug 18, 2026

Uh oh!

ssaketh-ch commented Aug 18, 2026

Uh oh!

psap-forge-bot Bot commented Aug 18, 2026

🟢 Execution of rhaiis nvidia benchmark hera ci-quick 🟢

✅ 00 Pre-Cleanup 1 second

✅ 01 Prepare 2 seconds

✅ 02 Preflight 1 second

✅ 03 Test 5 minutes, 14 seconds

✅ 04 Post-Cleanup 4 seconds

🔄 05 Export-Artifacts

Uh oh!

psap-forge-bot Bot commented Aug 18, 2026

Uh oh!

ssaketh-ch commented Aug 18, 2026

Uh oh!

psap-forge-bot Bot commented Aug 18, 2026

🟢 Execution of rhaiis nvidia benchmark hera ci-quick 🟢

✅ 00 Pre-Cleanup 1 second

✅ 01 Prepare 2 seconds

✅ 02 Preflight 1 second

✅ 03 Test 5 minutes, 15 seconds

✅ 04 Post-Cleanup 4 seconds

🔄 05 Export-Artifacts

Uh oh!

psap-forge-bot Bot commented Aug 18, 2026

Uh oh!

albertoperdomo2 commented Aug 19, 2026

Uh oh!

albertoperdomo2 commented Aug 19, 2026

Uh oh!

openshift-ci Bot commented Aug 19, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

coderabbitai Bot commented Aug 17, 2026 •

edited

Loading

naveenmiriyaluredhat commented Aug 17, 2026 •

edited

Loading

🟢 Execution of `rhaiis` nvidia benchmark hera ci-quick 🟢

✅ 00 Pre-Cleanup `1 second`

✅ 01 Prepare `2 seconds`

✅ 02 Preflight `1 second`

✅ 03 Test `5 minutes, 14 seconds`

✅ 04 Post-Cleanup `4 seconds`

🟢 Execution of `rhaiis` nvidia benchmark hera ci-quick 🟢

✅ 00 Pre-Cleanup `1 second`

✅ 01 Prepare `2 seconds`

✅ 02 Preflight `1 second`

✅ 03 Test `5 minutes, 15 seconds`

✅ 04 Post-Cleanup `4 seconds`