fix typing errors

Author: voorhs

src/autointent/_dump_tools/unit_dumpers.py

@@ -228,7 +228,7 @@ def load(path: Path, **kwargs: Any) -> "PeftModel":  # noqa: ANN401, ARG004
             # prompt learning model
             ptuning_path = path / "ptuning"
             model = transformers.AutoModelForSequenceClassification.from_pretrained(ptuning_path / "base_model")
-            return peft.PeftModel.from_pretrained(model, ptuning_path / "peft")
+            return peft.PeftModel.from_pretrained(model, ptuning_path / "peft")  # type: ignore[no-any-return]
         if (path / "lora").exists():
             # merged lora model
             lora_path = path / "lora"
@@ -278,7 +278,7 @@ def dump(obj: "PreTrainedTokenizer | PreTrainedTokenizerFast", path: Path, exist
     @staticmethod
     def load(path: Path, **kwargs: Any) -> "PreTrainedTokenizer | PreTrainedTokenizerFast":  # noqa: ANN401, ARG004
         transformers = require("transformers", extra="transformers")
-        return transformers.AutoTokenizer.from_pretrained(path)  # type: ignore[no-any-return,no-untyped-call]
+        return transformers.AutoTokenizer.from_pretrained(path)  # type: ignore[no-any-return]
 
     @classmethod
     def check_isinstance(cls, obj: Any) -> bool:  # noqa: ANN401

src/autointent/generation/utterances/_adversarial/human_utterance_generator.py

@@ -114,7 +114,7 @@ def augment(
             generated_split = HFDataset.from_list(new_samples)
             dataset[split_name] = concatenate_datasets([original_split, generated_split])
 
-        return [Sample(**sample) for sample in new_samples]
+        return [Sample.model_validate(sample) for sample in new_samples]
 
     async def augment_async(
         self, dataset: Dataset, split_name: str = Split.TRAIN, update_split: bool = True, n_final_per_class: int = 5

src/autointent/modules/scoring/_bert.py

@@ -151,7 +151,7 @@ def fit(
     ) -> None:
         self._validate_task(labels)
 
-        self._tokenizer = self._AutoTokenizer.from_pretrained(self.classification_model_config.model_name)  # type: ignore[no-untyped-call]
+        self._tokenizer = self._AutoTokenizer.from_pretrained(self.classification_model_config.model_name)
         self._model = self._initialize_model()
         tokenized_dataset = self._get_tokenized_dataset(utterances, labels)
         self._train(tokenized_dataset)

src/autointent/modules/scoring/_gcn/gcn_model.py

@@ -90,7 +90,7 @@ def create_correlation_matrix(train_labels: torch.Tensor, num_classes: int, p: f
         reweighted_adj = adj_matrix_no_self_loop * weights_p.unsqueeze(1)
         reweighted_adj.fill_diagonal_(1 - p)
 
-        return cast(torch.Tensor, reweighted_adj)
+        return reweighted_adj
 
     def set_correlation_matrix(self, train_labels: torch.Tensor) -> None:
         corr_matrix = self.create_correlation_matrix(