RecursionError: maximum recursion depth exceeded

[DesiKeki]

Hi,

I am trying to use the expr.to('excel') utility of skompile, and I am getting following error:

RecursionError: maximum recursion depth exceeded

The error stack is as below:

---------------------------------------------------------------------------
RecursionError                            Traceback (most recent call last)
<ipython-input-150-7dcc4a57e7a5> in <module>
      1 expr = skompile(final_model.predict)
      2 print(expr)
----> 3 expr.to('excel')
      4 #expr.to('sqlalchemy/sqlite')

/remote/vghawk1/nayaka/soft/anaconda3/envs/tf2_pytorch/lib/python3.7/site-packages/skompiler/ast.py in to(self, target, *args, **kw)
    266             mod = import_module(module)
    267             translator = getattr(mod, callable)
--> 268         return translator(self, *dialect, *args, **kw)
    269 #endregion
    270 

/remote/vghawk1/nayaka/soft/anaconda3/envs/tf2_pytorch/lib/python3.7/site-packages/skompiler/fromskast/excel.py in translate(node, component, multistage, assign_to, multistage_subexpression_min_length, _max_subexpression_length)
     62                          multistage_subexpression_min_length=multistage_subexpression_min_length,
     63                          _max_subexpression_length=_max_subexpression_length)
---> 64     result = writer(node)
     65     if component is not None:
     66         result = result[component]

/remote/vghawk1/nayaka/soft/anaconda3/envs/tf2_pytorch/lib/python3.7/site-packages/skompiler/fromskast/_common.py in __call__(self, node, **kw)
     31 
     32     def __call__(self, node, **kw):
---> 33         return getattr(self, node.__class__.__name__)(node, **kw)
     34 
     35 

/remote/vghawk1/nayaka/soft/anaconda3/envs/tf2_pytorch/lib/python3.7/site-packages/skompiler/fromskast/_common.py in LFold(self, node, **kw)
     80         if not node.elems:
     81             raise ValueError("LFold expects at least one element")
---> 82         return self(reduce(lambda x, y: BinOp(node.op, x, y), node.elems), **kw)
     83 
     84     def Let(self, node, **kw):

/remote/vghawk1/nayaka/soft/anaconda3/envs/tf2_pytorch/lib/python3.7/site-packages/skompiler/fromskast/_common.py in __call__(self, node, **kw)
     31 
     32     def __call__(self, node, **kw):
---> 33         return getattr(self, node.__class__.__name__)(node, **kw)
     34 
     35 

/remote/vghawk1/nayaka/soft/anaconda3/envs/tf2_pytorch/lib/python3.7/site-packages/skompiler/fromskast/_common.py in BinOp(self, node, **kw)
     64         the operation elementwise and returns a list."""
     65 
---> 66         left = self(node.left, **kw)
     67         right = self(node.right, **kw)
     68         op = self(node.op, **kw)

... last 2 frames repeated, from the frame below ...

/remote/vghawk1/nayaka/soft/anaconda3/envs/tf2_pytorch/lib/python3.7/site-packages/skompiler/fromskast/_common.py in __call__(self, node, **kw)
     31 
     32     def __call__(self, node, **kw):
---> 33         return getattr(self, node.__class__.__name__)(node, **kw)
     34 
     35 

RecursionError: maximum recursion depth exceeded

Please help in resolving this issue.

Regards
Keki

[DesiKeki]

Update:
Got same error for expr.to('sqlalchemy/sqlite')

[konstantint]

Try doing

import sys
sys.setrecursionlimit(10000)

before the expr.to(...) call.

[DesiKeki]

Thanks a lot @konstantint
That fixed the error. But I am not sure if I am correctly using SKompiler for my purpose. It will be a great help if you could guide me for this.

In my project I am using a RandomForestRegressor to predict certain outputs with given inputs. My aim is to check and analyse (preferably visually) the contribution/weights of my model's input features which are governing a particular output prediction. Something like what we can do in a neural network: (https://playground.tensorflow.org/).

Can SKompiler help me do that? If not, do you by chance have any idea how can I achieve my purpose?

[konstantint]

I think you should instead check out treeinterpreter.

[DesiKeki]

Thanks @konstantint
I had already started working with treeinterpreter. But I have multi-label predictions and with treeinterpreter I am getting following message:

**ValueError**: Multilabel classification trees not supported

I am trying to work this around. I'll close this thread now as the original doubt put in this thread is resolved. Thanks a lot for your help :).

Regards
Keki