fix(_bert): pin revision on transformers v5 PEFT adapter probe

When PEFT is installed, transformers v5 calls find_adapter_config_file
on every AutoModelForSequenceClassification.from_pretrained. The
auto_factory only propagates `_commit_hash` (used for the cache
lookup) but NOT the outer `revision` to the fall-through
hf_hub_download. On a cold cache — i.e. our CI warm-cache job, which
populates model files but no negative marker for adapter_config.json —
that probe fires `hf_hub_download(repo_id, adapter_config.json,
revision=None)` and our test guard rightly flagged it as unpinned.

Pass `adapter_kwargs={"revision": revision}` so the adapter probe
inherits the pin. The first run still writes a `.no_exist` marker, but
all subsequent runs (and CI's pinned-only contract) stay clean.

Reproduces with: rm -rf ~/.cache/huggingface/hub/models--prajjwal1--bert-tiny/.no_exist
then pytest tests/pipeline/test_inference.py::test_inference_from_config[multiclass].

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

Author: voorhs

src/autointent/modules/scoring/_bert.py

@@ -136,14 +136,21 @@ def _initialize_model(self) -> Any:  # noqa: ANN401
         label2id = {str(i): i for i in range(self._n_classes)}
         id2label = {i: str(i) for i in range(self._n_classes)}
 
+        # transformers v5 + PEFT triggers find_adapter_config_file on every
+        # from_pretrained; it propagates _commit_hash for the cache lookup but
+        # NOT the outer `revision` to the fall-through hf_hub_download
+        # (auto_factory.py:308 only forwards adapter_kwargs). Set revision
+        # explicitly via adapter_kwargs so the adapter probe stays pinned.
+        revision = self.classification_model_config.revision
         return AutoModelForSequenceClassification.from_pretrained(
             self.classification_model_config.model_name,
             trust_remote_code=self.classification_model_config.trust_remote_code,
-            revision=self.classification_model_config.revision,
+            revision=revision,
             num_labels=self._n_classes,
             label2id=label2id,
             id2label=id2label,
             problem_type="multi_label_classification" if self._multilabel else "single_label_classification",
+            adapter_kwargs={"revision": revision} if revision is not None else None,
         )
 
     def fit(