google-research/timesfm

Largest clones in the code

• 130 lines in 2 placespython1,056 tokens
  src/timesfm/utils/xreg_lib.py:72–201·v1/src/timesfm/xreg_lib.py:56–182

  72class BatchedInContextXRegBase:
  73  """Helper class for in-context regression covariate formatting.
  74
  75  Attributes:
  76    targets: List of targets (responses) of the in-context regression.
  77    train_lens: List of lengths of each target vector from the context.
  78    test_lens: List of lengths of each forecast horizon.
  79    train_dynamic_numerical_covariates: Dict of covariate names mapping to the
  80      dynamic numerical covariates of each forecast task on the context. Their
  81      lengths should match the corresponding lengths in `train_lens`.
  

  First 10 of 130 lines.

• 165 lines in 2 placespython777 tokens
  src/timesfm/utils/xreg_lib.py:207–371·v1/src/timesfm/xreg_lib.py:190–340

  207  self.static_numerical_covariates = static_numerical_covariates or {}
  208  self.static_categorical_covariates = static_categorical_covariates or {}
  209
  210def _assert_covariates(self, assert_covariate_shapes: bool = False) -> None:
  211  """Verifies the validity of the covariate inputs."""
  212
  213  # Check presence.
  214  if (
  215    self.train_dynamic_numerical_covariates
  216    and not self.test_dynamic_numerical_covariates
  

  190  self.static_numerical_covariates = static_numerical_covariates or {}
  191  self.static_categorical_covariates = static_categorical_covariates or {}
  192
  193def _assert_covariates(self, assert_covariate_shapes: bool = False) -> None:
  194  """Verifies the validity of the covariate inputs."""
  195
  196  # Check presence.
  197  if (self.train_dynamic_numerical_covariates and
  198      not self.test_dynamic_numerical_covariates) or (
  199          not self.train_dynamic_numerical_covariates and
  

  First 10 of 165 lines, the second copy on the right.

• 88 lines in 2 placespython626 tokens
  src/timesfm/utils/xreg_lib.py:377–464·v1/src/timesfm/xreg_lib.py:355–440

  377  x_test = [(x_test - x_mean) / x_std]
  378
  379# Categorical features. Encode one by one.
  380one_hot_encoder = preprocessing.OneHotEncoder(
  381  drop=one_hot_encoder_drop,
  382  sparse_output=False,
  383  handle_unknown="ignore",
  384)
  385for name in sorted(self.train_dynamic_categorical_covariates.keys()):
  386  ohe_train = _unnest(self.train_dynamic_categorical_covariates[name])[
  

  355  x_test = [np.concatenate(norm_test, axis=0)]
  356
  357# Categorical features. Encode one by one.
  358one_hot_encoder = preprocessing.OneHotEncoder(
  359    drop=one_hot_encoder_drop,
  360    sparse_output=False,
  361    handle_unknown="ignore",
  362)
  363for name in sorted(self.train_dynamic_categorical_covariates.keys()):
  364  ohe_train = _unnest(
  

  First 10 of 88 lines, the second copy on the right.