Automated Code Change

PiperOrigin-RevId: 944116839

Author: h-joo

lit_nlp/api/components.py

@@ -137,7 +137,7 @@ def is_compatible(self, model: lit_model.Model,
     """True if the model and dataset support metric computation."""
     for pred_spec in model.output_spec().values():
       parent_key: Optional[str] = getattr(pred_spec, 'parent', None)
-      parent_spec: Optional[types.LitType] = dataset.spec().get(parent_key)
+      parent_spec: Optional[types.LitType] = dataset.spec().get(parent_key)  # pyrefly: ignore[bad-argument-type]
       if self.is_field_compatible(pred_spec, parent_spec):
         return True
     return False

lit_nlp/api/dataset.py

@@ -236,7 +236,7 @@ def remap(self, field_map: Mapping[str, str]):
     """Return a copy of this dataset with some fields renamed."""
     new_spec = utils.remap_dict(self.spec(), field_map)
     new_examples = [utils.remap_dict(ex, field_map) for ex in self.examples]
-    return Dataset(new_spec, new_examples, base=self)
+    return Dataset(new_spec, new_examples, base=self)  # pyrefly: ignore[bad-argument-type]
 
   @staticmethod
   def lit_example_from_bytes(input_bytes: bytes) -> Optional[JsonDict]:
@@ -273,7 +273,7 @@ def index_inputs(
       indexed.append(
           IndexedInput(
               data=types.MappingProxyType(
-                  example | {INPUT_ID_FIELD: ex_id, INPUT_META_FIELD: ex_meta}
+                  example | {INPUT_ID_FIELD: ex_id, INPUT_META_FIELD: ex_meta}  # pyrefly: ignore[unsupported-operation]
               ),
               id=ex_id,
               meta=ex_meta,
@@ -376,6 +376,7 @@ def load(self, path: str):
       new_dataset = base_dataset.load(path) if base_dataset else None
 
       if new_dataset is not None:
+        # pyrefly: ignore[missing-attribute]
         description = (f'{len(new_dataset)} examples from '
                        f'{path}\n{self._base.description()}')
         return IndexedDataset(
@@ -438,7 +439,7 @@ def load_lit_format(
         **kw,
     )
   else:
-    return Dataset(spec=spec, examples=examples, *args, **kw)
+    return Dataset(spec=spec, examples=examples, *args, **kw)  # pyrefly: ignore[bad-argument-type]
 
 
 # TODO(b/202210900): Remove "NoneDataset" once the LIT front-end constructs its

lit_nlp/api/layout.py

@@ -70,7 +70,7 @@ def __call__(self, **kw):
 # LINT.IfChange
 # TODO(lit-dev): consider making modules subclass this instead of LitModuleName.
 @attr.s(auto_attribs=True)
-class ModuleConfig(dtypes.DataTuple):
+class ModuleConfig(dtypes.DataTuple):  # pyrefly: ignore[invalid-inheritance]
   module: Union[str, LitModuleName]
   requiredForTab: bool = False
   # TODO(b/172979677): support title, duplicateAsRow, numCols,
@@ -88,15 +88,15 @@ class ModuleConfig(dtypes.DataTuple):
 
 
 @attr.s(auto_attribs=True)
-class LayoutSettings(dtypes.DataTuple):
+class LayoutSettings(dtypes.DataTuple):  # pyrefly: ignore[invalid-inheritance]
   hideToolbar: bool = False
   mainHeight: int = 45
   leftWidth: int = 50
   centerPage: bool = False
 
 
 @attr.s(auto_attribs=True)
-class LitCanonicalLayout(dtypes.DataTuple):
+class LitCanonicalLayout(dtypes.DataTuple):  # pyrefly: ignore[invalid-inheritance]
   """Frontend UI layout; should match client/lib/types.ts."""
   upper: LitTabGroupLayout
   lower: LitTabGroupLayout = attr.ib(factory=dict)

lit_nlp/api/model.py

@@ -178,12 +178,12 @@ def load(self, path: str):
   @abc.abstractmethod
   def input_spec(self) -> types.Spec:
     """Return a spec describing model inputs."""
-    return
+    return  # pyrefly: ignore[bad-return]
 
   @abc.abstractmethod
   def output_spec(self) -> types.Spec:
     """Return a spec describing model outputs."""
-    return
+    return  # pyrefly: ignore[bad-return]
 
   def get_embedding_table(self) -> tuple[list[str], np.ndarray]:
     """Return the full vocabulary and embedding table.
@@ -369,7 +369,7 @@ def predict_minibatch(self, inputs: list[JsonDict]) -> list[JsonDict]:
     Returns:
       list of outputs, following model.output_spec()
     """
-    return
+    return  # pyrefly: ignore[bad-return]
 
 
 class ProjectorModel(BatchedModel, metaclass=abc.ABCMeta):

lit_nlp/api/types.py

@@ -155,10 +155,10 @@ def from_json(d: JsonDict):
 
     base_cls = globals().get("LitType")
     cls = globals().get(type_name)  # class by name from this module
-    if cls is None or not issubclass(cls, base_cls):
+    if cls is None or not issubclass(cls, base_cls):  # pyrefly: ignore[bad-argument-type]
       raise NameError(f"{type_name} is not a valid LitType.")
 
-    return cls(**{k: d[k] for k in d if k != "__name__"})
+    return cls(**{k: d[k] for k in d if k != "__name__"})  # pyrefly: ignore[not-callable]
 
 Spec = dict[str, LitType]
 
@@ -174,7 +174,7 @@ def _remap_leaf(leaf: LitType, keymap: dict[str, str]) -> LitType:
       k: (keymap.get(v, v) if k in FIELD_REF_ATTRIBUTES else v)
       for k, v in d.items()
   }
-  return leaf.__class__(**d)
+  return leaf.__class__(**d)  # pyrefly: ignore[bad-argument-type]
 
 
 def remap_spec(spec: Spec, keymap: dict[str, str]) -> Spec:
@@ -201,7 +201,7 @@ class StringLitType(LitType):
   Mainly used for string inputs that have special formatting, and should only
   be edited manually.
   """
-  default: str = ""
+  default: str = ""  # pyrefly: ignore[bad-override]
 
   def validate_input(self, value, spec: Spec, example: Input):
     if not isinstance(value, str):
@@ -260,13 +260,13 @@ def validate_output(self, value, output_spec: Spec, output_dict: JsonDict,
 @attr.s(auto_attribs=True, frozen=True, kw_only=True)
 class ListLitType(LitType):
   """List type."""
-  default: Sequence[Any] = None
+  default: Sequence[Any] = None  # pyrefly: ignore[bad-assignment, bad-override]
 
 
 @attr.s(auto_attribs=True, frozen=True, kw_only=True)
 class _StringCandidateList(ListLitType):
   """A list of (text, score) tuples."""
-  default: ScoredTextCandidates = None
+  default: ScoredTextCandidates = None  # pyrefly: ignore[bad-assignment]
 
   def validate_output(self, value, output_spec: Spec, output_dict: JsonDict,
                       input_spec: Spec, dataset_spec: Spec,
@@ -373,9 +373,9 @@ class TokenTopKPreds(ListLitType):
 
   The inner list should contain (word, probability) in descending order.
   """
-  default: Sequence[ScoredTextCandidates] = None
+  default: Sequence[ScoredTextCandidates] = None  # pyrefly: ignore[bad-assignment]
 
-  align: str = None  # name of a Tokens field in the model output
+  align: str = None  # name of a Tokens field in the model output  # pyrefly: ignore[bad-assignment]
   parent: Optional[str] = None
 
   def _validate_scored_candidates(self, scored_candidates):
@@ -389,7 +389,7 @@ def _validate_scored_candidates(self, scored_candidates):
       if scored_candidate[1] is not None:
         if not isinstance(scored_candidate[1], NumericTypes):
           raise ValueError(f"{scored_candidate} second element is not a num")
-        if prev_val < scored_candidate[1]:
+        if prev_val < scored_candidate[1]:  # pyrefly: ignore[unsupported-operation]
           raise ValueError(
               "TokenTopKPreds candidates are not in descending order")
         else:
@@ -414,7 +414,7 @@ class Scalar(LitType):
   """Scalar value, a single float or int."""
   min_val: float = 0
   max_val: float = 1
-  default: float = 0
+  default: float = 0  # pyrefly: ignore[bad-override]
   step: float = .01
 
   def validate_input(self, value, spec: Spec, example: Input):
@@ -464,7 +464,7 @@ class TokenScores(_FloatList):
 @attr.s(auto_attribs=True, frozen=True, kw_only=True)
 class ReferenceScores(ListLitType):
   """Score of one or more target sequences."""
-  default: Sequence[float] = None
+  default: Sequence[float] = None  # pyrefly: ignore[bad-assignment]
 
   # name of a TextSegment or ReferenceTexts field in the input
   parent: Optional[str] = None
@@ -504,7 +504,7 @@ def validate_input(self, value, spec: Spec, example: Input):
 @attr.s(auto_attribs=True, frozen=True, kw_only=True)
 class _Tensor(LitType):
   """A tensor type."""
-  default: Sequence[float] = None
+  default: Sequence[float] = None  # pyrefly: ignore[bad-assignment, bad-override]
 
   def validate_input(self, value, spec: Spec, example: Input):
     if isinstance(value, list):
@@ -585,7 +585,7 @@ class SpanLabels(ListLitType):
   Span labels can cover more than one token, may not cover all tokens in the
   sentence, and may overlap with each other.
   """
-  default: Sequence[dtypes.SpanLabel] = None
+  default: Sequence[dtypes.SpanLabel] = None  # pyrefly: ignore[bad-assignment]
   align: str  # name of Tokens field
   parent: Optional[str] = None
 
@@ -609,7 +609,7 @@ class EdgeLabels(ListLitType):
   https://github.com/nyu-mll/jiant/tree/master/probing#data-format for more
   details.
   """
-  default: Sequence[dtypes.EdgeLabel] = None
+  default: Sequence[dtypes.EdgeLabel] = None  # pyrefly: ignore[bad-assignment]
   align: str  # name of Tokens field
 
   def validate_output(self, value, output_spec: Spec, output_dict: JsonDict,
@@ -636,7 +636,7 @@ class MultiSegmentAnnotations(ListLitType):
   TODO(lit-dev): by default, spans are treated as bytes in this context.
   Make this configurable, if some spans need to refer to tokens instead.
   """
-  default: Sequence[dtypes.AnnotationCluster] = None
+  default: Sequence[dtypes.AnnotationCluster] = None  # pyrefly: ignore[bad-assignment]
   exclusive: bool = False  # if true, treat as candidate list
   background: bool = False  # if true, don't emphasize in visualization
 
@@ -775,7 +775,7 @@ class SubwordOffsets(ListLitType):
 
   offsets[i] should be the index of the first wordpiece for input token i.
   """
-  default: Sequence[int] = None
+  default: Sequence[int] = None  # pyrefly: ignore[bad-assignment]
   align_in: str  # name of field in data spec
   align_out: str  # name of field in model output spec
 
@@ -793,7 +793,7 @@ class SparseMultilabelPreds(_StringCandidateList):
 
   The tuples are of the label and the score.
   """
-  default: ScoredTextCandidates = None
+  default: ScoredTextCandidates = None  # pyrefly: ignore[bad-assignment]
   vocab: Optional[Sequence[str]] = None  # label names
   parent: Optional[str] = None
 
@@ -816,7 +816,7 @@ class SingleFieldMatcher(FieldMatcher):
 
   UI will materialize this to a dropdown-list.
   """
-  default: str = None
+  default: str = None  # pyrefly: ignore[bad-assignment, bad-override]
 
 
 @attr.s(auto_attribs=True, frozen=True, kw_only=True)
@@ -826,7 +826,7 @@ class MultiFieldMatcher(FieldMatcher):
   UI will materialize this to multiple checkboxes. Use this when the user needs
   to pick more than one field in UI.
   """
-  default: Sequence[str] = []  # default names of selected items.
+  default: Sequence[str] = []  # default names of selected items.  # pyrefly: ignore[bad-override]
   select_all: bool = False  # Select all by default (overriddes default).
 
 
@@ -840,20 +840,20 @@ class Salience(LitType):
 @attr.s(auto_attribs=True, frozen=True, kw_only=True)
 class TokenSalience(Salience):
   """Metadata about a returned token salience map."""
-  default: dtypes.TokenSalience = None
+  default: dtypes.TokenSalience = None  # pyrefly: ignore[bad-assignment, bad-override]
 
 
 @attr.s(auto_attribs=True, frozen=True, kw_only=True)
 class FeatureSalience(Salience):
   """Metadata about a returned feature salience map."""
-  default: dtypes.FeatureSalience = None
+  default: dtypes.FeatureSalience = None  # pyrefly: ignore[bad-assignment, bad-override]
 
 
 @attr.s(auto_attribs=True, frozen=True, kw_only=True)
 class FrameSalience(Salience):
   """Metadata about a returned frame salience map."""
 
-  default: dtypes.FrameSalience = None
+  default: dtypes.FrameSalience = None  # pyrefly: ignore[bad-assignment, bad-override]
 
 
 @attr.s(auto_attribs=True, frozen=True, kw_only=True)
@@ -869,13 +869,13 @@ class ImageSalience(Salience):
 @attr.s(auto_attribs=True, frozen=True, kw_only=True)
 class SequenceSalience(Salience):
   """Metadata about a returned sequence salience map."""
-  default: dtypes.SequenceSalienceMap = None
+  default: dtypes.SequenceSalienceMap = None  # pyrefly: ignore[bad-assignment, bad-override]
 
 
 @attr.s(auto_attribs=True, frozen=True, kw_only=True)
 class BooleanLitType(LitType):
   """Boolean value."""
-  default: bool = False
+  default: bool = False  # pyrefly: ignore[bad-override]
 
   def validate_input(self, value, spec, example: Input):
     if not isinstance(value, bool):
@@ -916,7 +916,7 @@ class MetricBestValue(dtypes.EnumSerializableAsValues, enum.Enum):
 @attr.s(auto_attribs=True, frozen=True, kw_only=True)
 class MetricResult(LitType):
   """Score returned from the computation of a Metric."""
-  default: float = 0
+  default: float = 0  # pyrefly: ignore[bad-override]
   description: str = ""
   best_value: MetricBestValue = MetricBestValue.NONE
 
@@ -934,7 +934,7 @@ class SalienceTargetInfo(LitType):
 
   Value is a dict with keys 'field' (str) and 'index' (Optional[int]).
   """
-  default: Optional[Mapping[str, Any]] = None
+  default: Optional[Mapping[str, Any]] = None  # pyrefly: ignore[bad-override]
 
 
 # LINT.ThenChange(../client/lib/lit_types.ts)

lit_nlp/app.py

@@ -111,7 +111,7 @@ def _build_metadata(self):
       }
 
       # List compatible datasets.
-      info['datasets'] = [
+      info['datasets'] = [  # pyrefly: ignore[bad-assignment]
           name for name, dataset in self._datasets.items()
           if model.is_compatible_with_dataset(dataset)
       ]
@@ -134,13 +134,13 @@ def _build_metadata(self):
             _get_compatible_names(self._metrics, model, dataset)
         )
 
-      info['generators'] = [
+      info['generators'] = [  # pyrefly: ignore[bad-assignment]
           name for name in self._generators.keys() if name in compat_gens
       ]
-      info['interpreters'] = [
+      info['interpreters'] = [  # pyrefly: ignore[bad-assignment]
           name for name in self._interpreters.keys() if name in compat_interps
       ]
-      info['metrics'] = [
+      info['metrics'] = [  # pyrefly: ignore[bad-assignment]
           name for name in self._metrics.keys() if name in compat_metrics
       ]
       model_info[name] = info
@@ -213,7 +213,7 @@ def _reconstitute_inputs(
         len(inputs),
         dataset_name,
     )
-    return [index[ex] if isinstance(ex, str) else ex for ex in inputs]
+    return [index[ex] if isinstance(ex, str) else ex for ex in inputs]  # pyrefly: ignore[bad-index]
 
   def _save_datapoints(
       self,
@@ -297,16 +297,16 @@ def _get_preds(self,
 
     # Figure out what to return to the frontend.
     output_spec = self._get_model_spec(model)['output']
-    requested_types = requested_types.split(',') if requested_types else []
-    requested_fields = requested_fields.split(',') if requested_fields else []
+    requested_types = requested_types.split(',') if requested_types else []  # pyrefly: ignore[bad-assignment]
+    requested_fields = requested_fields.split(',') if requested_fields else []  # pyrefly: ignore[bad-assignment]
     logging.info('Requested types: %s, fields: %s', str(requested_types),
                  str(requested_fields))
-    for t_name in requested_types:
+    for t_name in requested_types:  # pyrefly: ignore[not-iterable]
       t_class = getattr(types, t_name, None)
-      if not issubclass(t_class, types.LitType):
+      if not issubclass(t_class, types.LitType):  # pyrefly: ignore[bad-argument-type]
         raise TypeError(f"Class '{t_name}' is not a valid LitType.")
       requested_fields.extend(utils.find_spec_keys(output_spec, t_class))
-    ret_keys = set(requested_fields)  # de-dupe
+    ret_keys = set(requested_fields)  # de-dupe  # pyrefly: ignore[bad-argument-type]
 
     # Return selected keys.
     logging.info('Will return keys: %s', str(ret_keys))
@@ -861,7 +861,7 @@ def _run_annotators(self,
     for annotator in self._annotators:
       annotator.annotate(datapoints, dataset, annotated_spec)
     return lit_dataset.Dataset(
-        base=dataset, examples=datapoints, spec=annotated_spec)
+        base=dataset, examples=datapoints, spec=annotated_spec)  # pyrefly: ignore[bad-argument-type]
 
   def make_handler(self, fn):
     """Convenience wrapper to handle args and serialization.
@@ -1009,9 +1009,9 @@ def __init__(
 
     # Interpreter initialization
     if interpreters is not None:
-      self._interpreters = core.required_interpreters() | interpreters
+      self._interpreters = core.required_interpreters() | interpreters  # pyrefly: ignore[unsupported-operation]
     else:
-      self._interpreters = core.default_interpreters(self._models)
+      self._interpreters = core.default_interpreters(self._models)  # pyrefly: ignore[bad-argument-type]
 
     if metrics is not None:
       self._metrics = metrics