Bump pip from 25.2 to 25.3

atomgen/data/data_collator.py

@@ -81,8 +81,8 @@ def torch_call(
         # Handle dict or lists with proper padding and conversion to tensor.
         if self.pad:
             if isinstance(examples[0], Mapping):
-                batch: Dict[str, Any] = self.tokenizer.pad(  # type: ignore[assignment]
-                    examples,  # type: ignore[arg-type]
+                batch: Dict[str, Any] = self.tokenizer.pad(
+                    examples,
                     return_tensors="pt",
                     pad_to_multiple_of=self.pad_to_multiple_of,
                 )
@@ -186,7 +186,7 @@ def torch_mask_tokens(
         inputs = torch.where(
             ~mask,
             inputs,
-            self.tokenizer.convert_tokens_to_ids(self.tokenizer.mask_token),  # type: ignore[arg-type]
+            self.tokenizer.convert_tokens_to_ids(self.tokenizer.mask_token),
         )
         labels = torch.where(mask, labels, -100)
         if special_tokens_mask is not None:

atomgen/data/tokenizer.py

@@ -13,7 +13,7 @@
 VOCAB_FILES_NAMES: Dict[str, str] = {"vocab_file": "tokenizer.json"}
 
 
-class AtomTokenizer(PreTrainedTokenizer):
+class AtomTokenizer(PreTrainedTokenizer):  # type: ignore[misc]
     """
     Tokenizer for atomistic data.
 
@@ -43,7 +43,7 @@ def __init__(
             [(ids, tok) for tok, ids in self.vocab.items()]
         )
 
-        super().__init__(  # type: ignore[no-untyped-call]
+        super().__init__(
             pad_token=pad_token,
             mask_token=mask_token,
             bos_token=bos_token,
@@ -63,7 +63,7 @@ def load_vocab(vocab_file: str) -> Dict[str, int]:
                 )
         return vocab
 
-    def _tokenize(self, text: str) -> List[str]:  # type: ignore[override]
+    def _tokenize(self, text: str) -> List[str]:
         """Tokenize the text."""
         tokens = []
         i = 0
@@ -95,7 +95,7 @@ def convert_tokens_to_string(self, tokens: List[str]) -> str:
         """Convert the list of chemical symbol tokens to a concatenated string."""
         return "".join(tokens)
 
-    def pad(  # type: ignore[override]
+    def pad(
         self,
         encoded_inputs: Union[
             BatchEncoding,
@@ -155,7 +155,7 @@ def pad(  # type: ignore[override]
                     pad_to_multiple_of=pad_to_multiple_of,
                 )
 
-        return super().pad(
+        return super().pad(  # type: ignore[no-any-return]
             encoded_inputs=encoded_inputs,
             padding=padding,
             max_length=max_length,

atomgen/models/modeling_atomformer.py

@@ -2550,7 +2550,7 @@ def forward(
 class AtomformerPreTrainedModel(PreTrainedModel):  # type: ignore[no-untyped-call]
     """Base class for all transformer models."""
 
-    config_class = AtomformerConfig
+    config_class = AtomformerConfig  # type: ignore[assignment]
     base_model_prefix = "model"
     supports_gradient_checkpointing = True
     _no_split_modules = ["ParallelBlock"]

atomgen/models/schnet.py

@@ -134,7 +134,7 @@ class SchNetPreTrainedModel(PreTrainedModel):  # type: ignore[no-untyped-call]
     simple interface for loading and exporting models.
     """
 
-    config_class = SchNetConfig
+    config_class = SchNetConfig  # type: ignore[assignment]
     base_model_prefix = "model"
     supports_gradient_checkpointing = False
 

atomgen/models/tokengt.py

@@ -2510,7 +2510,7 @@ def custom_forward(*inputs: Any) -> Any:
 class TransformerPreTrainedModel(PreTrainedModel):  # type: ignore[no-untyped-call]
     """Base class for all transformer models."""
 
-    config_class = TransformerConfig
+    config_class = TransformerConfig  # type: ignore[assignment]
     base_model_prefix = "model"
     supports_gradient_checkpointing = True
     _no_split_modules = ["ParallelBlock"]