[BUG] Unable to save TapNet estimators because of Lambda layers #3539
Description
Describe the bug
The underlying keras model of TapNetClassifier and TapNetRegressor cannot be saved using model.save() because of Lambda layers in the implementation.
To Reproduce
import numpy as np
from sktime.classification.deep_learning.tapnet import TapNetClassifier
X = np.random.randn(15, 28, 16)
y = np.random.randint(0, high=2, size=(15, ))
tapnet = TapNetClassifier()
tapnet.fit(X, y)
tapnet.save("tapnet")Error Trace:
...
File "D:\Anaconda\envs\sktime-dev\lib\site-packages\keras\utils\traceback_utils.py", line 67, in error_handler
raise e.with_traceback(filtered_tb) from None
File "D:\Anaconda\envs\sktime-dev\lib\copy.py", line 146, in deepcopy
y = copier(x, memo)
File "D:\Anaconda\envs\sktime-dev\lib\copy.py", line 230, in _deepcopy_dict
y[deepcopy(key, memo)] = deepcopy(value, memo)
File "D:\Anaconda\envs\sktime-dev\lib\copy.py", line 146, in deepcopy
y = copier(x, memo)
File "D:\Anaconda\envs\sktime-dev\lib\copy.py", line 205, in _deepcopy_list
append(deepcopy(a, memo))
File "D:\Anaconda\envs\sktime-dev\lib\copy.py", line 172, in deepcopy
y = _reconstruct(x, memo, *rv)
File "D:\Anaconda\envs\sktime-dev\lib\copy.py", line 296, in _reconstruct
value = deepcopy(value, memo)
File "D:\Anaconda\envs\sktime-dev\lib\copy.py", line 146, in deepcopy
y = copier(x, memo)
File "D:\Anaconda\envs\sktime-dev\lib\copy.py", line 230, in _deepcopy_dict
y[deepcopy(key, memo)] = deepcopy(value, memo)
File "D:\Anaconda\envs\sktime-dev\lib\copy.py", line 146, in deepcopy
y = copier(x, memo)
File "D:\Anaconda\envs\sktime-dev\lib\copy.py", line 210, in _deepcopy_tuple
y = [deepcopy(a, memo) for a in x]
File "D:\Anaconda\envs\sktime-dev\lib\copy.py", line 210, in <listcomp>
y = [deepcopy(a, memo) for a in x]
File "D:\Anaconda\envs\sktime-dev\lib\copy.py", line 146, in deepcopy
y = copier(x, memo)
File "D:\Anaconda\envs\sktime-dev\lib\copy.py", line 210, in _deepcopy_tuple
y = [deepcopy(a, memo) for a in x]
File "D:\Anaconda\envs\sktime-dev\lib\copy.py", line 210, in <listcomp>
y = [deepcopy(a, memo) for a in x]
File "D:\Anaconda\envs\sktime-dev\lib\copy.py", line 161, in deepcopy
rv = reductor(4)
TypeError: cannot pickle 'module' objectAdditional context
This issue was known during migration and needs to be worked upon. The solution is subclassing tf.keras.layers.Layer to perform the same operation that Lambda layers currently do.