Automated Code Change

PiperOrigin-RevId: 943252298

Author: h-joo

python/dp_auditorium/dp_auditorium/examples/run_mean_mechanism_example.py

@@ -82,7 +82,7 @@ def mean_mechanism_report(
       max_value=1.0,
   )
   # Initialize the mechanism.
-  mechanism = mean.MeanMechanism(mech_config, rng)
+  mechanism = mean.MeanMechanism(mech_config, rng)  # pyrefly: ignore[bad-argument-type]
 
   # Configuration for a Hockey-Stick property tester. Given arrays s1 and s2
   # with samples two distributions it will estimate the hockey-stick divergence
@@ -135,7 +135,7 @@ def mean_mechanism_report(
       property_tester=property_tester,
   )
 
-  return test_runner.test_privacy(mechanism, "non-private-mean-mechanism")
+  return test_runner.test_privacy(mechanism, "non-private-mean-mechanism")  # pyrefly: ignore[bad-argument-type]
 
 
 def main(argv: Sequence[str]) -> None:

python/dp_auditorium/dp_auditorium/mechanisms/mean.py

@@ -55,11 +55,11 @@ def __init__(
     if self._delta == 0.0:
       # Laplace mechanism.
       self._noise_factor = 1.0
-      self._noise_function = self._rng.laplace
+      self._noise_function = self._rng.laplace  # pyrefly: ignore[missing-attribute]
     else:
       # Gaussian mechanism.
       self._noise_factor = np.sqrt(2.0 * np.log(1.25 / self._delta))
-      self._noise_function = self._rng.normal
+      self._noise_function = self._rng.normal  # pyrefly: ignore[missing-attribute]
 
   def _get_counts(
       self, data: np.ndarray, num_samples: int
@@ -77,7 +77,7 @@ def _get_counts(
       the entry in the tuple is an array of length num_samples.
     """
     noise_scale = 1.0 / self._epsilon_budget_scale / self._epsilon
-    noisy_counts = len(data) + self._rng.laplace(
+    noisy_counts = len(data) + self._rng.laplace(  # pyrefly: ignore[missing-attribute]
         0, noise_scale, (num_samples, 1)
     )
     if self._use_noised_counts_for_calculating_mean:

python/dp_auditorium/dp_auditorium/mechanisms/noisy_max.py

@@ -48,7 +48,7 @@ def __call__(self, data: np.ndarray, num_samples: int) -> np.ndarray:
     counts = collections.Counter(data)
     counts_array = [counts[i] for i in range(self.num_elements)]
 
-    noise = self.rng.gumbel(
+    noise = self.rng.gumbel(  # pyrefly: ignore[missing-attribute]
         loc=0, scale=2 / self.epsilon, size=(num_samples, self.num_elements)
     )
     noisy_counts = counts_array + noise

python/dp_auditorium/dp_auditorium/mechanisms/pipeline_dp/aggregation.py

@@ -80,7 +80,7 @@ def _compute_aggregations(self, data: list[float]) -> list[float]:
   def __call__(self, data: np.ndarray, num_samples: int) -> np.ndarray:
     """Returns an array of samples of a DP aggregation using pipeline_dp."""
     result = []
-    data = list(data)  # PipelineDP works now for list only.
+    data = list(data)  # PipelineDP works now for list only.  # pyrefly: ignore[bad-assignment]
     for _ in range(num_samples):
-      result.append(self._compute_aggregations(data))
+      result.append(self._compute_aggregations(data))  # pyrefly: ignore[bad-argument-type]
     return np.array(result)

python/dp_auditorium/dp_auditorium/mechanisms/sparse_vector_technique.py

@@ -256,14 +256,14 @@ def get_noisy_queries(
         * self._queries.sensitivity
         / self._epsilon
     )
-    return queries_values + self._rng.laplace(
+    return queries_values + self._rng.laplace(  # pyrefly: ignore[missing-attribute]
         0, query_noise_scale, size=(num_samples, self._num_queries)
     )
 
   @override
   def get_noisy_thresholds(self, num_samples: int) -> np.ndarray:
     threshold_noise_scale = 2 * self._queries.sensitivity / self._epsilon
-    return self._threshold + self._rng.laplace(
+    return self._threshold + self._rng.laplace(  # pyrefly: ignore[missing-attribute]
         0, threshold_noise_scale, size=(num_samples, 1)
     )
 
@@ -273,7 +273,7 @@ def get_query_output(
   ) -> tuple[np.ndarray, np.ndarray]:
     queries_pass_threshold = noisy_queries >= noisy_threshold
     return (
-        np.where(
+        np.where(  # pyrefly: ignore[no-matching-overload]
             queries_pass_threshold,
             mechanism_config.DataValues.SVT_ABOVE_THRESHOLD,
             mechanism_config.DataValues.SVT_BELOW_THRESHOLD,
@@ -303,7 +303,7 @@ def get_noisy_queries(
         * self._queries.sensitivity
         / self._epsilon
     )
-    return queries_values + self._rng.laplace(
+    return queries_values + self._rng.laplace(  # pyrefly: ignore[missing-attribute]
         0, query_noise_scale, size=(num_samples, self._num_queries)
     )
 
@@ -315,7 +315,7 @@ def get_noisy_thresholds(self, num_samples: int) -> np.ndarray:
         * self._max_answered_queries
         / self._epsilon
     )
-    return self._threshold + self._rng.laplace(
+    return self._threshold + self._rng.laplace(  # pyrefly: ignore[missing-attribute]
         0,
         threshold_noise_scale,
         size=(num_samples, 1),
@@ -327,7 +327,7 @@ def get_query_output(
   ) -> tuple[np.ndarray, np.ndarray]:
     queries_pass_threshold = noisy_queries >= noisy_threshold
     return (
-        np.where(
+        np.where(  # pyrefly: ignore[no-matching-overload]
             queries_pass_threshold,
             mechanism_config.DataValues.SVT_ABOVE_THRESHOLD,
             mechanism_config.DataValues.SVT_BELOW_THRESHOLD,
@@ -359,14 +359,14 @@ def get_noisy_queries(
         * self._queries.sensitivity
         / self._epsilon
     )
-    return queries_values + self._rng.laplace(
+    return queries_values + self._rng.laplace(  # pyrefly: ignore[missing-attribute]
         0, query_noise_scale, size=(num_samples, self._num_queries)
     )
 
   @override
   def get_noisy_thresholds(self, num_samples: int) -> np.ndarray:
     threshold_noise_scale = 2 * self._queries.sensitivity / self._epsilon
-    return self._threshold + self._rng.laplace(
+    return self._threshold + self._rng.laplace(  # pyrefly: ignore[missing-attribute]
         0, threshold_noise_scale, size=(num_samples, 1)
     )
 
@@ -376,7 +376,7 @@ def get_query_output(
   ) -> tuple[np.ndarray, np.ndarray]:
     queries_pass_threshold = noisy_queries >= noisy_threshold
     return (
-        np.where(
+        np.where(  # pyrefly: ignore[no-matching-overload]
             queries_pass_threshold,
             noisy_queries,
             mechanism_config.DataValues.SVT_BELOW_THRESHOLD,
@@ -403,14 +403,14 @@ def get_noisy_queries(
       self, queries_values: np.ndarray, num_samples: int
   ) -> np.ndarray:
     query_noise_scale = 4 * self._queries.sensitivity / (3 * self._epsilon)
-    return queries_values + self._rng.laplace(
+    return queries_values + self._rng.laplace(  # pyrefly: ignore[missing-attribute]
         0, query_noise_scale, size=(num_samples, self._num_queries)
     )
 
   @override
   def get_noisy_thresholds(self, num_samples: int) -> np.ndarray:
     threshold_noise_scale = 4 * self._queries.sensitivity / self._epsilon
-    return self._threshold + self._rng.laplace(
+    return self._threshold + self._rng.laplace(  # pyrefly: ignore[missing-attribute]
         0,
         threshold_noise_scale,
         size=(num_samples, 1),
@@ -422,7 +422,7 @@ def get_query_output(
   ) -> tuple[np.ndarray, np.ndarray]:
     queries_pass_threshold = noisy_queries >= noisy_threshold
     return (
-        np.where(
+        np.where(  # pyrefly: ignore[no-matching-overload]
             queries_pass_threshold,
             mechanism_config.DataValues.SVT_ABOVE_THRESHOLD,
             mechanism_config.DataValues.SVT_BELOW_THRESHOLD,
@@ -451,7 +451,7 @@ def get_noisy_queries(
   @override
   def get_noisy_thresholds(self, num_samples: int) -> np.ndarray:
     threshold_noise_scale = 2 * self._queries.sensitivity / self._epsilon
-    return self._threshold + self._rng.laplace(
+    return self._threshold + self._rng.laplace(  # pyrefly: ignore[missing-attribute]
         0,
         threshold_noise_scale,
         size=(num_samples, 1),
@@ -463,7 +463,7 @@ def get_query_output(
   ) -> tuple[np.ndarray, np.ndarray]:
     queries_pass_threshold = noisy_queries >= noisy_threshold
     return (
-        np.where(
+        np.where(  # pyrefly: ignore[no-matching-overload]
             queries_pass_threshold,
             mechanism_config.DataValues.SVT_ABOVE_THRESHOLD,
             mechanism_config.DataValues.SVT_BELOW_THRESHOLD,
@@ -488,14 +488,14 @@ def get_noisy_queries(
       self, queries_values: np.ndarray, num_samples: int
   ) -> np.ndarray:
     query_noise_scale = 2 * self._queries.sensitivity / self._epsilon
-    return queries_values + self._rng.laplace(
+    return queries_values + self._rng.laplace(  # pyrefly: ignore[missing-attribute]
         0, query_noise_scale, size=(num_samples, self._num_queries)
     )
 
   @override
   def get_noisy_thresholds(self, num_samples: int) -> np.ndarray:
     threshold_noise_scale = 2 * self._queries.sensitivity / self._epsilon
-    return self._threshold + self._rng.laplace(
+    return self._threshold + self._rng.laplace(  # pyrefly: ignore[missing-attribute]
         0,
         threshold_noise_scale,
         size=(num_samples, 1),
@@ -507,7 +507,7 @@ def get_query_output(
   ) -> tuple[np.ndarray, np.ndarray]:
     queries_pass_threshold = noisy_queries >= noisy_threshold
     return (
-        np.where(
+        np.where(  # pyrefly: ignore[no-matching-overload]
             queries_pass_threshold,
             mechanism_config.DataValues.SVT_ABOVE_THRESHOLD,
             mechanism_config.DataValues.SVT_BELOW_THRESHOLD,

python/dp_auditorium/dp_auditorium/testers/histogram_tester.py

@@ -32,7 +32,7 @@
 
 
 def _estimate_discrete_distribution(
-    samples: IntArray,
+    samples: IntArray,  # pyrefly: ignore[not-a-type]
     universe_size: int,
 ) -> np.ndarray:
   """Returns estimated probability mass function over universe using samples.

python/dp_auditorium/dp_auditorium/testers/renyi_tester.py

@@ -80,7 +80,7 @@ def train_step(
       self, data: tuple[np.ndarray, np.ndarray]
   ) -> Dict[str, tf.Tensor]:
     with tf.GradientTape() as tape:
-      divergence = self(data, training=True)
+      divergence = self(data, training=True)  # pyrefly: ignore[not-callable]
       loss = -divergence
 
     trainable_vars = self.nn_model.trainable_variables
@@ -89,7 +89,7 @@ def train_step(
     return {'divergence': divergence}
 
   def call(  # pytype: disable=annotation-type-mismatch
-      self, data: tuple[np.ndarray, np.ndarray], training: bool = None
+      self, data: tuple[np.ndarray, np.ndarray], training: bool = None  # pyrefly: ignore[bad-function-definition]
   ) -> tf.Tensor:
     """Estimate renyi divergence from samples and current nn_model.
 
@@ -229,14 +229,14 @@ def _compute_divergence_on_samples(
       samples2_test: np.ndarray,
       failure_probability: float,
   ) -> float:
-    divergence_test = model((samples1_test, samples2_test))
+    divergence_test = model((samples1_test, samples2_test))  # pyrefly: ignore[not-callable]
 
     # Calculate lower end of confidence interval.
     num_samples = min(samples1_test.shape[0], samples2_test.shape[0])
     error = _compute_error_from_samples(
         num_samples=num_samples,
         failure_probability=failure_probability,
-        model_output_coordinate_bound=self._model_output_coordinate_bound,
+        model_output_coordinate_bound=self._model_output_coordinate_bound,  # pyrefly: ignore[bad-argument-type]
         alpha=self._alpha,
     )
     divergence_test_lower_bound = divergence_test - error