Skip to content

[DO NOT MERGE] control: hf-adapters suite on legacy runners (baseline for the /next shadow PR) - #183

Open
HarikrishnanBalagopal wants to merge 1 commit into
torch-spyre:mainfrom
HarikrishnanBalagopal:ci-cd/test
Open

[DO NOT MERGE] control: hf-adapters suite on legacy runners (baseline for the /next shadow PR)#183
HarikrishnanBalagopal wants to merge 1 commit into
torch-spyre:mainfrom
HarikrishnanBalagopal:ci-cd/test

Conversation

@HarikrishnanBalagopal

@HarikrishnanBalagopal HarikrishnanBalagopal commented Jul 15, 2026

Copy link
Copy Markdown
Collaborator

Control PR — DO NOT MERGE.

Purpose: answer one question with no confounds — does the hf-adapters test
suite pass on the legacy CI runner image, with the source unchanged?

  • Base main, reset clean. The only diff is a single marker file
    (.github/CONTROL-PR.md). No image redirect, no venv changes, no test
    edits. That keeps this a faithful baseline; the marker file (rather than an
    empty commit) forces a real regression run, since an empty commit can
    resolve the test-type oddly.
  • Legacy runners. Jobs land on the default image_torch_spyre (legacy
    2.0/ci image), exactly as main would.
  • Pairs with the /next-image shadow PR (same suite, /next runner
    image). Read them together: shadow red + this control green ⇒ the /next
    image is the cause, not the tests.

Fixes do not belong here. Any change to make tests pass goes in a dedicated
fix PR (e.g. the uv baked-venv fix in #509), never in this control. Earlier
revisions of this branch had smuggled-in venv fixes; those have been stripped
so the control is honest again.

@spyre-ci

spyre-ci Bot commented Sep 4, 2026

Copy link
Copy Markdown

🔄 spyre-test: running

Plan (build waves + dependencies, per arch)

amd64

flowchart LR
  subgraph Lamd64_0["amd64 L0 · 1 parallel"]
    n_amd64_hf_adapters_hf_adapters_dev["hf-adapters/hf-adapters-dev<br/>image · ec634f161831"]
  end
  subgraph Lamd64_1["amd64 L1 · 1 parallel"]
    n_amd64_spyre_inference_spyre_inference_dev["spyre-inference/spyre-inference-dev<br/>image · 49b2e8756eb0"]
  end
  n_amd64_hf_adapters_hf_adapters_dev --> n_amd64_spyre_inference_spyre_inference_dev
  classDef sPending fill:#eceff1,stroke:#90a4ae,color:#37474f
  classDef sBuilding fill:#fff8e1,stroke:#f9a825,color:#5d4037,stroke-width:2px
  classDef sOk fill:#e8f5e9,stroke:#43a047,color:#1b5e20
  classDef sReused fill:#e3f2fd,stroke:#1e88e5,color:#0d47a1
  classDef sFailed fill:#ffebee,stroke:#e53935,color:#b71c1c,stroke-width:2px
  classDef sDropped fill:#f5f5f5,stroke:#bdbdbd,color:#9e9e9e
  class n_amd64_hf_adapters_hf_adapters_dev sOk;
  class n_amd64_spyre_inference_spyre_inference_dev sOk;
Loading

🔨 orch trigger-pr-validation · arches amd64 · fp amd64=797daaf2

level component arch build smoke unit integration trunk regression perf
L0 hf-adapters/hf-adapters-dev amd64 ✅ ok · · · ·
L1 spyre-inference/spyre-inference-dev amd64 ✅ ok · · · ·

@spyre-ci

spyre-ci Bot commented Sep 4, 2026

Copy link
Copy Markdown

❌ spyre-test: failure

Plan (build waves + dependencies, per arch)

amd64

flowchart LR
  subgraph Lamd64_0["amd64 L0 · 1 parallel"]
    n_amd64_hf_adapters_hf_adapters_dev["hf-adapters/hf-adapters-dev<br/>image · ec634f161831"]
  end
  subgraph Lamd64_1["amd64 L1 · 1 parallel"]
    n_amd64_spyre_inference_spyre_inference_dev["spyre-inference/spyre-inference-dev<br/>image · 49b2e8756eb0"]
  end
  n_amd64_hf_adapters_hf_adapters_dev --> n_amd64_spyre_inference_spyre_inference_dev
Loading

HarikrishnanBalagopal added a commit to HarikrishnanBalagopal/hf-adapters that referenced this pull request Sep 9, 2026
torch-spyre#183)

The test steps located the venv only via $VIRTUAL_ENV, which the /next dev
image bakes+exports but the legacy prod-ci image does not. On the legacy runner
a bare `uv run --active` fell back to an empty project-local .venv, so pytest
was absent and 0 tests ran. Set UV_PROJECT_ENVIRONMENT at workflow level (the
same path build-hf-adapters populates) and fall back to that literal path in the
two source-activate sites, so the venv resolves the same on both images.

Signed-off-by: Harikrishnan Balagopal <harikrishmenon@gmail.com>
HarikrishnanBalagopal added a commit to HarikrishnanBalagopal/hf-adapters that referenced this pull request Sep 9, 2026
torch-spyre#183)

The test steps located the venv only via $VIRTUAL_ENV, which the /next dev
image bakes+exports but the legacy prod-ci image does not. On the legacy runner
a bare `uv run --active` fell back to an empty project-local .venv, so pytest
was absent and 0 tests ran. Fold UV_PROJECT_ENVIRONMENT into the workflow's
existing top-level env: block (the same path build-hf-adapters populates) and
fall back to that literal path in the two source-activate sites, so the venv
resolves the same on both images.

Signed-off-by: Harikrishnan Balagopal <harikrishmenon@gmail.com>
HarikrishnanBalagopal added a commit to HarikrishnanBalagopal/hf-adapters that referenced this pull request Sep 10, 2026
torch-spyre#183)

The test steps located the venv only via $VIRTUAL_ENV, which the /next dev
image bakes+exports but the legacy prod-ci image does not. On the legacy runner
a bare `uv run --active` fell back to an empty project-local .venv, so pytest
was absent and 0 tests ran. Fold UV_PROJECT_ENVIRONMENT into the workflow's
existing top-level env: block (the same path build-hf-adapters populates) and
fall back to that literal path in the two source-activate sites, so the venv
resolves the same on both images.

Signed-off-by: Harikrishnan Balagopal <harikrishmenon@gmail.com>
HarikrishnanBalagopal added a commit to HarikrishnanBalagopal/hf-adapters that referenced this pull request Sep 10, 2026
torch-spyre#183)

The test steps located the venv only via $VIRTUAL_ENV, which the /next dev
image bakes+exports but the legacy prod-ci image does not. On the legacy runner
a bare `uv run --active` fell back to an empty project-local .venv, so pytest
was absent and 0 tests ran. Fold UV_PROJECT_ENVIRONMENT into the workflow's
existing top-level env: block (the same path build-hf-adapters populates) and
fall back to that literal path in the two source-activate sites, so the venv
resolves the same on both images.

Signed-off-by: Harikrishnan Balagopal <harikrishmenon@gmail.com>
@HarikrishnanBalagopal
HarikrishnanBalagopal force-pushed the ci-cd/test branch 2 times, most recently from 4bed473 to f795378 Compare September 10, 2026 17:16
@HarikrishnanBalagopal HarikrishnanBalagopal changed the title ci-cd: DO NOT MERGE, only for triggering tests [DO NOT MERGE] control: hf-adapters suite on legacy runners (baseline for the /next shadow PR) Sep 10, 2026
Reset to main + a single marker file. No image redirect, no venv or test
changes. Runs the full suite on the default (legacy) image_torch_spyre
runners as a baseline for the /next-image shadow PR. Fixes go in dedicated
fix PRs, not here.

Signed-off-by: Harikrishnan Balagopal <harikrishmenon@gmail.com>
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.

1 participant