Skip to content

Ci test fixes#2079

Open
sebpop wants to merge 2 commits into
huggingface:mainfrom
sebpop:ci-test-fixes
Open

Ci test fixes#2079
sebpop wants to merge 2 commits into
huggingface:mainfrom
sebpop:ci-test-fixes

Conversation

@sebpop
Copy link
Copy Markdown
Contributor

@sebpop sebpop commented May 29, 2026

Two pre-existing CI failures on main, neither related to any code change in tokenizers/ or bindings/{python,node}/src/:

tests/python: use canonical Salesforce/wikitext dataset id — huggingface_hub enforces namespace/name for HF URIs; the bare id "wikitext" (which was relocated on the Hub) now produces HfUriError: Invalid HF URI 'hf://datasets/wikitext@/.huggingface.yaml'. Updates both datasets.load_dataset(...) call sites in test_tutorial_train_from_iterators.py.

tests/node: update bertProcessing error message to match napi-rs — napi-rs now reports "Failed to get Array length" for invalid array args (was "Given napi value is not an array"). Updates the assertion in post-processors.test.ts.

Each is a one-line / two-line change in tests only; no production code touched.

sebpop added 2 commits May 29, 2026 04:26
The bare dataset id "wikitext" has been relocated on the Hugging
Face Hub.  Newer huggingface_hub versions enforce the namespace/name
form and reject the bare id, so the @pytest.mark.network test
`test_tutorial_train_from_iterators::TestTrainFromIterators::test_datasets`
fails on every PR with:

  HfUriError: Invalid HF URI 'hf://datasets/wikitext@/.huggingface.yaml'.
  Repository id must be 'namespace/name'.

Update both `datasets.load_dataset(...)` call sites in this test to
use the canonical relocated id `Salesforce/wikitext`.  The config
name `wikitext-103-raw-v1` and split arguments are unchanged.
The napi-rs runtime now reports "Failed to get Array length" when an
expected `Array` argument is not an array, where it previously
reported "Given napi value is not an array".  The
`bindings/node/lib/bindings/post-processors.test.ts` test asserted
the older string, so it fails on every PR.

Update the expected error string to match the current napi-rs
behaviour.  The test still verifies that `bertProcessing(['sep', 1])`
(missing the second argument) throws; only the message text changes.
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