Fix syntax in EpsilonPlusFlat instantiation

[istepka]

There is a mismatch between the current implementation of composites here and in the Zennit core library. In two tutorial notebooks and in the README file, the EpsilonPlusFlat class is instantiated with incorrect arguments, which causes an error. This issue arises from passing the list of canonizers as a positional argument, whereas the source code clearly indicates it should be passed as a named argument.

I have fixed this issue in both notebooks and in the README.

Below is an actual example from the attribution.ipynb notebook.

from zennit.composites import EpsilonPlusFlat
from zennit.canonizers import SequentialMergeBatchNorm
from crp.attribution import CondAttribution

composite = EpsilonPlusFlat([SequentialMergeBatchNorm()])
attribution = CondAttribution(model, no_param_grad=True)

Output:

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
Cell In[16], [line 5](vscode-notebook-cell:?execution_count=16&line=5)
      [2](vscode-notebook-cell:?execution_count=16&line=2) from zennit.canonizers import SequentialMergeBatchNorm
      [3](vscode-notebook-cell:?execution_count=16&line=3) from crp.attribution import CondAttribution
----> [5](vscode-notebook-cell:?execution_count=16&line=5) composite = EpsilonPlusFlat([SequentialMergeBatchNorm()])
      [6](vscode-notebook-cell:?execution_count=16&line=6) attribution = CondAttribution(model, no_param_grad=True)

File ~/miniconda3/envs/lrp/lib/python3.12/site-packages/zennit/composites.py:425, in EpsilonPlusFlat.__init__(self, epsilon, stabilizer, layer_map, first_map, zero_params, canonizers)
    [420](https://file+.vscode-resource.vscode-cdn.net/home/ignacy/putai/thesis/forkzennit/zennit-crp/tutorials/~/miniconda3/envs/lrp/lib/python3.12/site-packages/zennit/composites.py:420)     first_map = []
    [422](https://file+.vscode-resource.vscode-cdn.net/home/ignacy/putai/thesis/forkzennit/zennit-crp/tutorials/~/miniconda3/envs/lrp/lib/python3.12/site-packages/zennit/composites.py:422) rule_kwargs = {'zero_params': zero_params}
    [423](https://file+.vscode-resource.vscode-cdn.net/home/ignacy/putai/thesis/forkzennit/zennit-crp/tutorials/~/miniconda3/envs/lrp/lib/python3.12/site-packages/zennit/composites.py:423) layer_map = layer_map + layer_map_base(stabilizer) + [
    [424](https://file+.vscode-resource.vscode-cdn.net/home/ignacy/putai/thesis/forkzennit/zennit-crp/tutorials/~/miniconda3/envs/lrp/lib/python3.12/site-packages/zennit/composites.py:424)     (Convolution, ZPlus(stabilizer=stabilizer, **rule_kwargs)),
--> [425](https://file+.vscode-resource.vscode-cdn.net/home/ignacy/putai/thesis/forkzennit/zennit-crp/tutorials/~/miniconda3/envs/lrp/lib/python3.12/site-packages/zennit/composites.py:425)     (torch.nn.Linear, Epsilon(epsilon=epsilon, **rule_kwargs)),
    [426](https://file+.vscode-resource.vscode-cdn.net/home/ignacy/putai/thesis/forkzennit/zennit-crp/tutorials/~/miniconda3/envs/lrp/lib/python3.12/site-packages/zennit/composites.py:426) ]
    [427](https://file+.vscode-resource.vscode-cdn.net/home/ignacy/putai/thesis/forkzennit/zennit-crp/tutorials/~/miniconda3/envs/lrp/lib/python3.12/site-packages/zennit/composites.py:427) first_map = first_map + [
    [428](https://file+.vscode-resource.vscode-cdn.net/home/ignacy/putai/thesis/forkzennit/zennit-crp/tutorials/~/miniconda3/envs/lrp/lib/python3.12/site-packages/zennit/composites.py:428)     (Linear, Flat(stabilizer=stabilizer, **rule_kwargs))
    [429](https://file+.vscode-resource.vscode-cdn.net/home/ignacy/putai/thesis/forkzennit/zennit-crp/tutorials/~/miniconda3/envs/lrp/lib/python3.12/site-packages/zennit/composites.py:429) ]
    [430](https://file+.vscode-resource.vscode-cdn.net/home/ignacy/putai/thesis/forkzennit/zennit-crp/tutorials/~/miniconda3/envs/lrp/lib/python3.12/site-packages/zennit/composites.py:430) super().__init__(layer_map=layer_map, first_map=first_map, canonizers=canonizers)

File ~/miniconda3/envs/lrp/lib/python3.12/site-packages/zennit/rules.py:115, in Epsilon.__init__(self, epsilon, zero_params)
    [114](https://file+.vscode-resource.vscode-cdn.net/home/ignacy/putai/thesis/forkzennit/zennit-crp/tutorials/~/miniconda3/envs/lrp/lib/python3.12/site-packages/zennit/rules.py:114) def __init__(self, epsilon=1e-6, zero_params=None):
--> [115](https://file+.vscode-resource.vscode-cdn.net/home/ignacy/putai/thesis/forkzennit/zennit-crp/tutorials/~/miniconda3/envs/lrp/lib/python3.12/site-packages/zennit/rules.py:115)     stabilizer_fn = Stabilizer.ensure(epsilon)
    [116](https://file+.vscode-resource.vscode-cdn.net/home/ignacy/putai/thesis/forkzennit/zennit-crp/tutorials/~/miniconda3/envs/lrp/lib/python3.12/site-packages/zennit/rules.py:116)     super().__init__(
    [117](https://file+.vscode-resource.vscode-cdn.net/home/ignacy/putai/thesis/forkzennit/zennit-crp/tutorials/~/miniconda3/envs/lrp/lib/python3.12/site-packages/zennit/rules.py:117)         input_modifiers=[lambda input: input],
    [118](https://file+.vscode-resource.vscode-cdn.net/home/ignacy/putai/thesis/forkzennit/zennit-crp/tutorials/~/miniconda3/envs/lrp/lib/python3.12/site-packages/zennit/rules.py:118)         param_modifiers=[NoMod(zero_params=zero_params)],
   (...)
    [121](https://file+.vscode-resource.vscode-cdn.net/home/ignacy/putai/thesis/forkzennit/zennit-crp/tutorials/~/miniconda3/envs/lrp/lib/python3.12/site-packages/zennit/rules.py:121)         reducer=(lambda inputs, gradients: inputs[0] * gradients[0]),
    [122](https://file+.vscode-resource.vscode-cdn.net/home/ignacy/putai/thesis/forkzennit/zennit-crp/tutorials/~/miniconda3/envs/lrp/lib/python3.12/site-packages/zennit/rules.py:122)     )

File ~/miniconda3/envs/lrp/lib/python3.12/site-packages/zennit/core.py:91, in Stabilizer.ensure(cls, value)
     [89](https://file+.vscode-resource.vscode-cdn.net/home/ignacy/putai/thesis/forkzennit/zennit-crp/tutorials/~/miniconda3/envs/lrp/lib/python3.12/site-packages/zennit/core.py:89) if callable(value):
     [90](https://file+.vscode-resource.vscode-cdn.net/home/ignacy/putai/thesis/forkzennit/zennit-crp/tutorials/~/miniconda3/envs/lrp/lib/python3.12/site-packages/zennit/core.py:90)     return value
---> [91](https://file+.vscode-resource.vscode-cdn.net/home/ignacy/putai/thesis/forkzennit/zennit-crp/tutorials/~/miniconda3/envs/lrp/lib/python3.12/site-packages/zennit/core.py:91) raise TypeError(f'Value {value} is not a valid stabilizer!')

TypeError: Value [<zennit.canonizers.SequentialMergeBatchNorm object at 0x7d0a5bd9bfe0>] is not a valid stabilizer!