[Python] Python Neural Network : TypeError: exceptions must derive from BaseException

I want to use python to training my neural network case, but I recognize a problem.

The console code looks like:



The original code:

@staticmethod
def colSpec(nparray: 'np.ndarray', rangeStr:str=None,batchStart:str=None, batchEnd:str=None)->'np.ndarray':
'''
Specify certain column data by selecting some columns or batched rows of data. --- UPDATED (Dexter) 20181210

Parameters
------------------------------

nparray `np.ndarray<np.array<(int|float|str)>>` - Data table to be selected.

rangeStr `str` - Index range string for selecting certain columns for the transformation.

batchStart `int` - The starting index (inclusive) of the data rows.

batchEnd `int` - The ending index (exclusive) of the data rows.

Returns
------------------------------

`np.ndarray<np.array<(int|float|str)>>` - Data table with requested selection range.
'''
# Convert to np array if it's not.
if not isinstance(nparray, np.ndarray):
if isinstance(nparray, list):
try:
nparray = np.array(nparray)
except:
raise ValueError("Data type not supported for colSpec. It should be an `np.ndarray` object.")
else:
raise ValueError("Data type not supported for colSpec. It should be an `np.ndarray` object.")

# If no column specification requested, just return the orginal array with batchStart and batchEnd rows.
if rangeStr is None:
return nparray[batchStart:batchEnd]

else:
colCount = len(nparray[0])

# Validate the index range range.
if (not IndexRange.validate(colCount, rangeStr)):
raise ValueError("Index range string is not valid.")

# Different scenarios on the column selections:
if ":" in rangeStr:
# Split the ":" to find the range start and end.
rangeInfo = [string.strip() for string in rangeStr.split(":")]

# None should be specified for 2-valued tuples.
if any([(ele == "None" or ele == "") for ele in rangeInfo]):
# Determine the scenarios for which None is used.
if rangeInfo[0] != "None" and len(rangeInfo[0]):
return nparray[batchStart:batchEnd,int(rangeInfo[0]):]
elif rangeInfo[1] != "None" and len(rangeInfo[1]):
return nparray[batchStart:batchEnd,:int(rangeInfo[1])]
else:
return nparray[batchStart:batchEnd]
else:
raise nparray[batchStart:batchEnd,int(rangeInfo[0]):int(rangeInfo[1])]

# Otherwise, it is specifying indexes explicitly.
else:
idxs = IndexRange.parse(colCount, rangeStr)
return nparray[batchStart:batchEnd,idxs]


I don't what's wrong about that, I even don't use exceptions code.

Continue reading...