Thanks to visit codestin.com
Credit goes to github.com

Skip to content

ci(windows): pull Gemma-4-E4B instead of legacy Llama-3.2-3B-Hybrid#1712

Merged
itomek merged 2 commits into
mainfrom
tmi/condescending-dhawan-e514c9
Jun 17, 2026
Merged

ci(windows): pull Gemma-4-E4B instead of legacy Llama-3.2-3B-Hybrid#1712
itomek merged 2 commits into
mainfrom
tmi/condescending-dhawan-e514c9

Conversation

@itomek

@itomek itomek commented Jun 17, 2026

Copy link
Copy Markdown
Collaborator

The Windows Agent SDK and CLI jobs were red because they pulled Llama-3.2-3B-Instruct-Hybrid — a legacy model whose download overran the 15-minute step budget and timed out the run before any test executed. This switches them to Gemma-4-E4B-it-GGUF (the repo's DEFAULT_MODEL_NAME) and pulls it through the running server's API instead of the lemonade CLI.

Why the API, not lemonade pull

The first attempt (lemonade pull Gemma-4-E4B-it-GGUF) failed on the 10.7.0 runners with When registering a new model, the model name must include the user namespace. Root cause: the model is a built-in of the C++ Lemonade server's catalog (recipe llamacpp, checkpoint unsloth/gemma-4-E4B-it-GGUF:Q4_K_M, present at the v10.7.0 tag), but the standalone lemonade CLI consults a separate registry that doesn't carry it. Routing the pull through POST /api/v1/pull on the already-healthy server — the same mechanism the green test_api.yml uses — hits the catalog that actually has the model, and the step fails loudly if the request errors.

Note: an earlier local "8/8 passed" check ran against Lemonade 10.2.0, where Gemma-4-E4B-it-GGUF was already cached/registered — warm-state masking that hid the cold-runner failure. Verification for this change is the 10.7.0 CI run on this PR.

Test plan

  • Test Agent SDK on Windows (Lemonade Integration) — model-pull step succeeds via the API and the Agent SDK integration tests run
  • Test GAIA CLI on Windows (Full Integration) — model-pull step succeeds and the Lemonade client integration tests run

The Agent SDK and CLI Windows jobs pulled Llama-3.2-3B-Instruct-Hybrid,
a legacy model whose download overran the 15-min step budget and timed
out the run. Switch to Gemma-4-E4B-it-GGUF (the repo DEFAULT_MODEL_NAME),
which the rest of the suite already provisions, and align the local-run
default in test_lemonade_client.py.
@github-actions github-actions Bot added devops DevOps/infrastructure changes tests Test changes labels Jun 17, 2026
@itomek itomek marked this pull request as ready for review June 17, 2026 23:27
@itomek itomek requested a review from kovtcharov-amd as a code owner June 17, 2026 23:27
@github-actions

Copy link
Copy Markdown
Contributor

Summary

Clean, correctly-scoped CI fix — approve. The two Windows jobs were timing out because they pulled Llama-3.2-3B-Instruct-Hybrid, whose download overran the 15-minute step budget before any test ran; swapping to Gemma-4-E4B-it-GGUF aligns CI with the repo's DEFAULT_MODEL_NAME (confirmed at src/gaia/llm/lemonade_client.py:119) so the suite actually executes. Eight additions, eight deletions, no logic changes. The single most important thing to know: the change is safe across all platforms because the only behavioral knob — the GAIA_TEST_MODEL default in tests/test_lemonade_client.py:59 — is overridden explicitly everywhere it matters, so this is a local-dev fallback bump, not a CI behavior change.

Issues Found

None blocking.

🟢 Legacy Llama-3.2-3B-Instruct-Hybrid still the default elsewhere (out of scope, informational)src/gaia/rag/app.py:251 still hardcodes it as the --model default, and the summarize README (src/gaia/apps/summarize/README.md:183,208) still uses it in examples. Not part of this PR's CI-timeout fix and correctly left untouched here, but worth a follow-up so user-facing defaults match DEFAULT_MODEL_NAME. No action needed on this PR.

Strengths

  • Verified from the real failure path, not a warm box. The description shows the actual pull step (the one that was timing out) completing in ~2.5 min plus the full 8/8 integration run — exactly the cold-state evidence CLAUDE.md asks for, rather than a mock that only proves the call was made.
  • Correctly scoped. Touches only the two red Windows workflows and the test default constant; no drive-by edits to the unrelated legacy references that still exist in the tree.
  • Consistency check passes. Both Windows jobs now pull and set GAIA_TEST_MODEL to the same Gemma model, and Linux CI pins its own model explicitly (Qwen3-0.6B-GGUF), so the changed default can't silently shift any CI job.

Verdict

Approve — no blocking issues; ready to merge once the two Windows jobs go green on the PR run.

@itomek itomek added this pull request to the merge queue Jun 17, 2026
Merged via the queue into main with commit e334b9e Jun 17, 2026
29 of 31 checks passed
@itomek itomek deleted the tmi/condescending-dhawan-e514c9 branch June 17, 2026 23:33
pull Bot pushed a commit to bhardwajRahul/gaia that referenced this pull request Jun 18, 2026
Follow-up to amd#1712, which auto-merged at its first commit and left
`main` with a raw `lemonade pull Gemma-4-E4B-it-GGUF` in both Windows
integration jobs. On the runners that 400s — `When registering a new
model, the model name must include the user namespace` — so the Agent
SDK and GAIA CLI Windows jobs fail at the model-pull step before any
test runs.

This routes the pull through `POST /api/v1/pull` on the already-running
server — the same mechanism the green `test_api.yml` uses — and fails
loudly if the request errors.

### Root cause
`Gemma-4-E4B-it-GGUF` is a genuine built-in of the Lemonade **server's**
catalog at v10.7.0 (recipe `llamacpp`). The failure is not the model or
the version — on a clean 10.7.0 install the CLI pull even works. The
runner's failure log showed **two** lemonade installs on `PATH`
(`...\systemprofile\...` and `C:\Users\user\...`); `lemonade pull`
resolved to one whose catalog doesn't carry Gemma-4, so it fell into the
"register a new model" path. Pulling through the running server's API
removes the dependency on whichever `lemonade` CLI happens to be first
on `PATH`.

### Verified on a 10.7.0 box (version parity with CI)
- `POST /api/v1/pull {model_name: Gemma-4-E4B-it-GGUF}` →
`{"status":"success"}`
- Gemma-4-E4B loads on 10.7.0's llama.cpp (AMD Radeon) → `Model loaded:
Gemma-4-E4B-it-GGUF`
- `test_agent_sdk.py` (GAIA_TEST_MODEL=Gemma-4-E4B-it-GGUF) → `Ran 8
tests in 28.761s / OK`

## Test plan
- [ ] `Test Agent SDK on Windows (Lemonade Integration)` — model-pull
step succeeds via the API and the integration tests run
- [ ] `Test GAIA CLI on Windows (Full Integration)` — model-pull step
succeeds and the tests run

Co-authored-by: Tomasz Iniewicz <[email protected]>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

devops DevOps/infrastructure changes tests Test changes

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants