[BUG] Getting NotImplementedError: <nvtabular.columns.selector.ColumnSelector object error from TargetEncoding op

[rnyak]

Describe the bug
Getting this error from TargetEncoding:

Failed to fit operator <nvtabular.ops.target_encoding.TargetEncoding object at 0x7f83cc8c9940>
Traceback (most recent call last):
  File "/nvtabular/nvtabular/workflow/workflow.py", line 220, in fit
    stats.append(op.fit(workflow_node.input_columns, transformed_ddf))
  File "/nvtabular/nvtabular/ops/target_encoding.py", line 169, in fit
    moments = _custom_moments(ddf[self.target])
  File "/root/.local/lib/python3.8/site-packages/dask-2021.7.1-py3.8.egg/dask/dataframe/core.py", line 4016, in __getitem__
    raise NotImplementedError(key)
NotImplementedError: <nvtabular.columns.selector.ColumnSelector object at 0x7f83cc894250>
---------------------------------------------------------------------------
NotImplementedError                       Traceback (most recent call last)
/tmp/ipykernel_8111/2721983559.py in <module>
     18 
     19 workflow = nvt.Workflow(te_features + ["Author", "Engaging_User"])
---> 20 df_out = workflow.fit_transform(nvt.Dataset(df)).to_ddf().compute(scheduler="synchronous")

/nvtabular/nvtabular/workflow/workflow.py in fit_transform(self, dataset)
    264         Dataset
    265         """
--> 266         self.fit(dataset)
    267         return self.transform(dataset)
    268 

/nvtabular/nvtabular/workflow/workflow.py in fit(self, dataset)
    218                 op = workflow_node.op
    219                 try:
--> 220                     stats.append(op.fit(workflow_node.input_columns, transformed_ddf))
    221                     ops.append(op)
    222                 except Exception:

/nvtabular/nvtabular/ops/target_encoding.py in fit(self, col_selector, ddf)
    167         if self.target_mean is None:
    168             # calcualte the mean if we don't have it already
--> 169             moments = _custom_moments(ddf[self.target])
    170 
    171         col_groups = col_selector.grouped_names

~/.local/lib/python3.8/site-packages/dask-2021.7.1-py3.8.egg/dask/dataframe/core.py in __getitem__(self, key)
   4014             return self.where(key, np.nan)
   4015 
-> 4016         raise NotImplementedError(key)
   4017 
   4018     def __setitem__(self, key, value):

NotImplementedError: <nvtabular.columns.selector.ColumnSelector object at 0x7f83cc894250>

Please run the following code to repro the issue

Steps/Code to reproduce bug

from nvtabular import ColumnSelector
import cudf
df = cudf.DataFrame({
        "Cost": range(15),
        "Post": [1, 2, 3, 4, 5] * 3,
        "Author": ['A'] * 5 + ['B'] * 5 + ['C'] * 2 + ['D'] * 3,
        "Engaging_User": ['A'] * 5 + ['B'] * 3 + ['E'] * 2 + ['D'] * 3 + ['G'] * 2})

cat_groups = ['Author', 'Engaging_User']
labels = ColumnSelector(['Post']) >> (lambda col: (col>3).astype('int8'))
te_features = cat_groups >> ops.TargetEncoding(
    labels,
    out_path='./',
    kfold=1,
    out_dtype="float32",
    drop_folds=False,  # Keep folds to validate
)

workflow = nvt.Workflow(te_features + ["Author", "Engaging_User"])
df_out = workflow.fit_transform(nvt.Dataset(df)).to_ddf().compute(scheduler="synchronous")

Expected behavior
TargetEncoding should be able to take a transformed workflow node as target.

Environment details (please complete the following information):

• Environment location: [Bare-metal, Docker, Cloud(specify cloud provider)]
• Method of NVTabular install: [conda, Docker, or from source]
  • If method of install is [Docker], provide docker pull & docker run commands used

Docker image: NVTabular-Pytorch-Training:21.09

[benfred]

This works in v0.6.0 (changing ColumnSelector -> ColumnGroup above), but is broken in v0.7.0

[albert17]

Then it looks like we are not covering TargetEncoding op properly with the Unit testing.