Added abstract functions for LM and distmod

[rohanmodi2810]

Description

Converted the LM and distmod functions in the doc to abstract functions in the library. I've added the calculate_probability() and sample() functions for the LanguageModel class. I've created a new class called DistributionModifier and added functions for each operators.

References

• Part of Convert the LM and distmod functions in the doc to abstract functions in the library #2

[neubig]

Hey @rohanmodi2810 , thanks a lot for doing this! Can I suggest an alternative structure?

class DistributionModifier:
    def __call__(base: LanguageModel, **kwargs):
        ...

class Fitter(DistributionModifier):
    def __call__(base: LanguageModel, target: LanguageModel, **kwargs):
        ...

...

etc.

So that each type of distmod function is a child class of distmod itself.

In addition, could we put all of these functions in the llments/distmod directory instead of llments/lm?

[neubig]

Superseded by #15, which we merged in already.