Dataset readers#

There are a set of classes to conveniently process datasets that exist in files:

These classes allow you to iterate the contents of the file as parsed examples. The reader adapts automatically based off if the Workspace is for single line or multiline inputs.

`TextFormatReader`#

import vowpal_wabbit_next as vw

workspace = vw.Workspace()

with open("example.txt", "r") as text_file:
    with vw.TextFormatReader(workspace, text_file) as reader:
        for example in reader:
            print(workspace.predict_then_learn_one(example))

0.0
0.14874061942100525
0.008019516244530678
-0.09955745935440063
-0.23391607403755188

`DSJsonFormatReader`#

import vowpal_wabbit_next as vw

workspace = vw.Workspace(["--cb_explore_adf"])

with open("example.dsjson", "r") as dsjson_file:
    with vw.DSJsonFormatReader(workspace, dsjson_file) as reader:
        for example in reader:
            print(workspace.predict_then_learn_one(example))

[(0, 0.25), (1, 0.25), (2, 0.25), (3, 0.25)]
[(0, 0.25), (1, 0.25), (2, 0.25), (3, 0.25)]
[(0, 0.9624999761581421), (3, 0.012500000186264515), (1, 0.012500000186264515), (2, 0.012500000186264515)]
[(0, 0.9624999761581421), (3, 0.012500000186264515), (1, 0.012500000186264515), (2, 0.012500000186264515)]
[(0, 0.9624999761581421), (3, 0.012500000186264515), (1, 0.012500000186264515), (2, 0.012500000186264515)]

`CacheFormatReader`#

See Cache format for an example of using the cache format.