[FEA] Support cuDF dataframes with Decimal dtype columns as input to estimators

[beckernick]

As cuDF continues to implement support for Decimal columns, we should accept Decimal columns as part of dataframe inputs to cuML estimators.

In PySpark, we can do this with the following:

from pyspark.sql import SparkSession
from pyspark.ml.classification import LogisticRegression
from pyspark.sql.types import DecimalType
from pyspark.ml.feature import VectorAssembler
​
spark = SparkSession.builder \
    .master("local") \
    .getOrCreate()
​
df = (spark
  .createDataFrame([
      [0.2,1,2.2,5,1],
      [0.3,2,3.2,4,0],
      [0.9,2,4.2,4,1]
  ], schema=["a","b","c","d", "label"])
     )
​
df = (df
      .withColumn('a', df['a'].cast('decimal(3,2)'))
      .withColumn('b', df['b'].cast('decimal(3,2)'))
      .withColumn('c', df['c'].cast('decimal(3,2)'))
      .withColumn('d', df['d'].cast('decimal(3,2)'))
)
​
assembler = VectorAssembler(
    inputCols=["a", "b", "c", "d"],
    outputCol="features")
​
training = assembler.transform(df)
​
lr = LogisticRegression(maxIter=10, regParam=0.3, elasticNetParam=0.8)
lrModel = lr.fit(training)
print("Coefficients: " + str(lrModel.coefficients))
print("Intercept: " + str(lrModel.intercept))
Coefficients: [0.0,0.0,0.0,0.0]
Intercept: 0.6931471805599453

import cudf
import cuml
import pyarrow as pa
​
df = cudf.DataFrame({
    "a":[0.1,0.2,0.9],
    "b":[0.3,0.9,20],
    "c":[-3.2, 0.2, 9.2],
    "label":[1,1,0]
})
​
pa_decimal_b = df.b.to_arrow().cast(pa.decimal128(precision=7, scale=2))
df["b"] = cudf.core.column.DecimalColumn.from_arrow(pa_decimal_b)
​
clf = cuml.linear_model.LogisticRegression()
clf.fit(df[["a","b","c"]], df["label"])
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-32-3b21064a47b7> in <module>
     16 
     17 clf = cuml.linear_model.LogisticRegression()
---> 18 clf.fit(df[["a","b","c"]], df["label"])

/raid/nicholasb/miniconda3/envs/rapids-gpu-bdb-automated-tests/lib/python3.7/site-packages/cuml/internals/api_decorators.py in inner_with_setters(*args, **kwargs)
    408                                 target_val=target_val)
    409 
--> 410                 return func(*args, **kwargs)
    411 
    412         @wraps(func)

cuml/linear_model/logistic_regression.pyx in cuml.linear_model.logistic_regression.LogisticRegression.fit()

/raid/nicholasb/miniconda3/envs/rapids-gpu-bdb-automated-tests/lib/python3.7/site-packages/cuml/internals/api_decorators.py in inner_with_setters(*args, **kwargs)
    408                                 target_val=target_val)
    409 
--> 410                 return func(*args, **kwargs)
    411 
    412         @wraps(func)

cuml/solvers/qn.pyx in cuml.solvers.qn.QN.fit()

/raid/nicholasb/miniconda3/envs/rapids-gpu-bdb-automated-tests/lib/python3.7/site-packages/cuml/internals/api_decorators.py in inner(*args, **kwargs)
    359         def inner(*args, **kwargs):
    360             with self._recreate_cm(func, args):
--> 361                 return func(*args, **kwargs)
    362 
    363         return inner

/raid/nicholasb/miniconda3/envs/rapids-gpu-bdb-automated-tests/lib/python3.7/site-packages/cuml/common/input_utils.py in input_to_cuml_array(X, order, deepcopy, check_dtype, convert_to_dtype, check_cols, check_rows, fail_on_order, force_contiguous)
    313             X_m = CumlArray(data=X.as_gpu_matrix(order='F'))
    314         else:
--> 315             X_m = CumlArray(data=X.as_gpu_matrix(order=order))
    316 
    317     elif isinstance(X, CumlArray):

/raid/nicholasb/miniconda3/envs/rapids-gpu-bdb-automated-tests/lib/python3.7/site-packages/cudf/core/dataframe.py in as_gpu_matrix(self, columns, order)
   3545         if any(
   3546             (is_categorical_dtype(c) or np.issubdtype(c, np.dtype("object")))
-> 3547             for c in cols
   3548         ):
   3549             raise TypeError("non-numeric data not yet supported")

/raid/nicholasb/miniconda3/envs/rapids-gpu-bdb-automated-tests/lib/python3.7/site-packages/cudf/core/dataframe.py in <genexpr>(.0)
   3545         if any(
   3546             (is_categorical_dtype(c) or np.issubdtype(c, np.dtype("object")))
-> 3547             for c in cols
   3548         ):
   3549             raise TypeError("non-numeric data not yet supported")

/raid/nicholasb/miniconda3/envs/rapids-gpu-bdb-automated-tests/lib/python3.7/site-packages/numpy/core/numerictypes.py in issubdtype(arg1, arg2)
    386     """
    387     if not issubclass_(arg1, generic):
--> 388         arg1 = dtype(arg1).type
    389     if not issubclass_(arg2, generic):
    390         arg2 = dtype(arg2).type

TypeError: Cannot interpret '<cudf.core.column.decimal.DecimalColumn object at 0x7f1ca5310170>' as a data type

[beckernick]

The underlying error actually comes from cuDF, but using this issue to track the ability to accept these inputs in cuML

[github-actions[bot]]

This issue has been labeled inactive-30d due to no recent activity in the past 30 days. Please close this issue if no further response or action is needed. Otherwise, please respond with a comment indicating any updates or changes to the original issue and/or confirm this issue still needs to be addressed. This issue will be labeled inactive-90d if there is no activity in the next 60 days.

[github-actions[bot]]

This issue has been labeled inactive-90d due to no recent activity in the past 90 days. Please close this issue if no further response or action is needed. Otherwise, please respond with a comment indicating any updates or changes to the original issue and/or confirm this issue still needs to be addressed.