Skip to main content

​
class EvalTable

columns: 'list[str | int] | None' = None,
data: 'list[Iterable[Any]] | np.ndarray | pd.DataFrame | None' = None,
rows: 'list[Iterable[Any]] | None' = None,
dataframe: 'pd.DataFrame | None' = None,
dtype: 'Any' = None,
optional: 'bool | list[bool]' = True,
allow_mixed_types: 'bool' = False,
log_mode: "Literal['IMMUTABLE']" = 'IMMUTABLE', *,
input_columns: 'list[str] | None' = None,
output_columns: 'list[str] | None' = None,
score_columns: 'list[str] | None' = None,
unsupported_media_mode: 'media_adapters.UnsupportedMediaMode' = 'stub'

​
Description

A Table subclass that routes run.log() to the new Eval Tables experience. When logged via run.log(), an EvalTable is logged as a Weave Eval via weave.EvaluationLogger instead of being uploaded as a regular wandb Table artifact. Note: EvalTable is a work-in-progress and is NOT yet officially released or supported.

​
Args

  • columns: Names of the columns in the table. If unset, but input_columns, output_columns, or score_columns are set, then we’ll just set columns to the union of those, in that order.
  • data:
  • rows:
  • dataframe:
  • dtype:
  • optional:
  • allow_mixed_types:
  • log_mode: Controls how the table is logged when the same EvalTable is passed to run.log() more than once.
  • input_columns: Names of the input columns. If set, designates these columns as inputs. Eval comparisons will match rows based on matching values from input columns. If unset, we will inject a “row” index input column so comparisons can match against that.
  • output_columns: Names of the output columns. These represents the values to be compared. Any columns not designated as input, output, or score will default to being output columns.
  • score_columns: Names of the score columns. These represent derived scores for the outputs. By default, we will auto-summarize any numeric and boolean scores.
  • unsupported_media_mode: How to handle unsupported wandb media/value types.

​
Methods

​
method add_column

self,
name: 'str | int',
data: 'list[Any] | np.ndarray',
optional: 'bool' = False
Adds a column of data to the table.
Arguments
  • name: (str) - the unique name of the column
  • data: (list | np.array) - a column of homogeneous data
  • optional: (bool) - if null-like values are permitted

​
method add_computed_columns

(self, fn)
Adds one or more computed columns based on existing data.
Arguments
  • fn: A function which accepts one or two parameters, ndx (int) and row (dict), which is expected to return a dict representing new columns for that row, keyed by the new column names. - ndx is an integer representing the index of the row. Only included if include_ndx is set to True. - row is a dictionary keyed by existing columns

​
method add_data

self, *data: 'Any'
Adds a new row of data to the table. The maximum amount ofrows in a table is determined by wandb.Table.MAX_ARTIFACT_ROWS. The length of the data should match the length of the table column.
Arguments
  • data:

​
method add_row

(self, *row)
Deprecated. Use Table.add_data method instead.
Arguments
  • row:

​
method cast

(self, col_name, dtype, optional=False)
Casts a column to a specific data type. This can be one of the normal python classes, an internal W&B type, or an example object, like an instance of wandb.Image or wandb.Classes.
Arguments
  • col_name: The name of the column to cast.
  • dtype: The target dtype.
  • optional: If the column should allow Nones.

​
method get_column

(self, name, convert_to=None)
Retrieves a column from the table and optionally converts it to a NumPy object.
Arguments
  • name: (str) - the name of the column
  • convert_to: (str, optional)

​
method get_dataframe

(self)
Returns a pandas.DataFrame of the table.

​
method get_index

(self)
Returns an array of row indexes for use in other tables to create links.