markovify model export benchmarks

speedtests_newclass_final.py

import markovify


# modified markovify Text class
class EditedTextClass(markovify.Text):
    def __init__(self, input_text, state_size=2, chain=None, runs=None):
        """
        input_text: A string.
        state_size: An integer, indicating the number of words in the model's state.
        chain: A trained markovify.Chain instance for this text, if pre-processed.
        """
        self.input_text = input_text
        self.state_size = state_size

        self.runs = runs or list(self.generate_corpus(self.input_text))

        # Rejoined text lets us assess the novelty of generated setences
        self.rejoined_text = self.sentence_join(map(self.word_join, self.runs))

        self.chain = chain or markovify.Chain(self.runs, state_size)

    def to_dict(self):
        return {
            "input_text": self.input_text,
            "state_size": self.state_size,
            "chain": self.chain.to_json(),
            "runs": self.runs
        }

    @classmethod
    def from_dict(cls, obj):
        return cls(
            obj["input_text"],
            state_size=obj["state_size"],
            chain=markovify.Chain.from_json(obj["chain"]),
            runs=obj["runs"]
        )


# Load test corpus to use for tests
with open('sherlock.txt', 'r', encoding='utf-8') as f:
    corpus = f.read()

# Create text model and get the generated model & chain JSONs to use in tests
# Standard markovify:
text_model = markovify.Text(corpus, state_size=2, chain=None)
chain_json_ss2 = text_model.chain.to_json()
model_json_ss2 = text_model.to_json()

# Modified markovify:
modified_text_model = EditedTextClass(corpus, state_size=2, chain=None)
modifiedmarkov_chain_json_ss2 = modified_text_model.chain.to_json()
modifiedmarkov_model_json_ss2 = modified_text_model.to_json()


# test functions for modified markovify class
def modifiedmarkovify_existing_model_size_2():
    return EditedTextClass.from_json(modifiedmarkov_model_json_ss2)


def modifiedmarkovify_existing_chain_size_2():
    return EditedTextClass.from_chain(modifiedmarkov_chain_json_ss2, corpus=corpus)


def modifiedmarkovify_new_model_size_2():
    return EditedTextClass(corpus, state_size=2, chain=None)


# test functions for current markovify

# Using existing model JSON
def existing_model_size_2():
    return markovify.Text.from_json(model_json_ss2)


# Using existing chain JSON
def existing_chain_size_2():
    return markovify.Text.from_chain(chain_json_ss2, corpus=corpus)


# From scratch
def new_model_size_2():
    return markovify.Text(corpus, state_size=2, chain=None)


if __name__ == '__main__':
    import timeit

    number_tests = 25
    print('Running {} times each, returning the average runtime in seconds.\n'.format(number_tests))

    print('Current markovify, create new model, state size 2:')
    print(timeit.timeit("new_model_size_2()", setup="from __main__ import new_model_size_2", number=number_tests) / number_tests)
    print('Example: {}'.format(new_model_size_2().make_short_sentence(120)))
    print('----------------------------------')

    print('Current markovify, existing chain JSON, state size 2:')
    print(timeit.timeit("existing_chain_size_2()", setup="from __main__ import existing_chain_size_2", number=number_tests) / number_tests)
    print('Example: {}'.format(existing_chain_size_2().make_short_sentence(120)))
    print('----------------------------------')

    print('Current markovify, existing model JSON, state size 2:')
    print(timeit.timeit("existing_model_size_2()", setup="from __main__ import existing_model_size_2", number=number_tests) / number_tests)
    print('Example: {}'.format(existing_model_size_2().make_short_sentence(120)))
    print('----------------------------------')

    print('\n\n\n')

    print('Modified markovify, create new model, state size 2:')
    print(timeit.timeit("modifiedmarkovify_new_model_size_2()", setup="from __main__ import modifiedmarkovify_new_model_size_2", number=number_tests) / number_tests)
    print('Example: {}'.format(modifiedmarkovify_new_model_size_2().make_short_sentence(120)))
    print('----------------------------------')

    print('Modified markovify, existing chain JSON, state size 2:')
    print(timeit.timeit("modifiedmarkovify_existing_chain_size_2()", setup="from __main__ import modifiedmarkovify_existing_chain_size_2", number=number_tests) / number_tests)
    print('Example: {}'.format(modifiedmarkovify_existing_chain_size_2().make_short_sentence(120)))
    print('----------------------------------')

    print('Modified markovify, existing model JSON, state size 2:')
    print(timeit.timeit("modifiedmarkovify_existing_model_size_2()", setup="from __main__ import modifiedmarkovify_existing_model_size_2", number=number_tests) / number_tests)
    print('Example: {}'.format(modifiedmarkovify_existing_model_size_2().make_short_sentence(120)))
    print('----------------------------------')

speedtests_newclass_profiled.py

import cProfile

import speedtests_newclass_final as speedtests_newclass


if __name__ == '__main__':
    print('Current markovify, create new model, state size 2:')
    pr1 = cProfile.Profile()
    pr1.enable()
    text_model = speedtests_newclass.new_model_size_2()
    pr1.disable()
    pr1.dump_stats('current_markovify_create_new_model_ss2.stats')
    print('Example: {}'.format(text_model.make_sentence()))
    print('----------------------------------')

    print('Current markovify, existing chain JSON, state size 2:')
    pr2 = cProfile.Profile()
    pr2.enable()
    text_model = speedtests_newclass.existing_chain_size_2()
    pr2.disable()
    pr2.dump_stats('current_markovify_existing_chain_json_ss2.stats')
    print('Example: {}'.format(text_model.make_sentence()))
    print('----------------------------------')

    print('Current markovify, existing model JSON, state size 2:')
    pr3 = cProfile.Profile()
    pr3.enable()
    text_model = speedtests_newclass.existing_model_size_2()
    pr3.disable()
    pr3.dump_stats('current_markovify_existing_model_ss2.stats')
    print('Example: {}'.format(text_model.make_sentence()))
    print('----------------------------------')

    print('\n\n\n')

    print('Modified markovify, create new model, state size 2:')
    pr7 = cProfile.Profile()
    pr7.enable()
    text_model = speedtests_newclass.modifiedmarkovify_new_model_size_2()
    pr7.disable()
    pr7.dump_stats('modified_markovify_new_model_ss2.stats')
    print('Example: {}'.format(text_model.make_sentence()))
    print('----------------------------------')

    print('Modified markovify, existing chain, state size 2:')
    pr8 = cProfile.Profile()
    pr8.enable()
    text_model = speedtests_newclass.modifiedmarkovify_existing_chain_size_2()
    pr8.disable()
    pr8.dump_stats('modified_markovify_existing_chain_ss2.stats')
    print('Example: {}'.format(text_model.make_sentence()))
    print('----------------------------------')

    print('Modified markovify, existing model, state size 2:')
    pr9 = cProfile.Profile()
    pr9.enable()
    text_model = speedtests_newclass.modifiedmarkovify_existing_model_size_2()
    pr9.disable()
    pr9.dump_stats('modified_markovify_existing_model_ss2.stats')
    print('Example: {}'.format(text_model.make_sentence()))
    print('----------------------------------')

timing tests with sherlock corpus (desktop).txt

Running 25 times each, returning the average runtime in seconds.

Current markovify, create new model, state size 2:
0.22077374984000925
Example: And so in ten minutes I had just transferred to it would give him an advance from a boot-lace.
----------------------------------
Current markovify, existing chain JSON, state size 2:
0.21371726716010017
Example: She was a man without heart or conscience.
----------------------------------
Current markovify, existing model JSON, state size 2:
0.23536155616006
Example: Posted to-day in Gravesend by a better fit if I did it, Mr. Holmes.
----------------------------------




Modified markovify, create new model, state size 2:
0.2151293533600983
Example: You took me to cut both ways.
----------------------------------
Modified markovify, existing chain JSON, state size 2:
0.2092507990000013
Example: The man sat huddled up in his conjecture, however, for he had come in, and the house.
----------------------------------
Modified markovify, existing model JSON, state size 2:
0.09925109259987948
Example: Your duties, as far as Reading, but could get the address of the journey, but the high, thin breathing of my dress.
----------------------------------

timing tests with sherlock corpus (raspi).txt

Running 25 times each, returning the average runtime in seconds.

Current markovify, create new model, state size 2:
16.18081657112576
Example: A blow was struck, and one small job, and that we should need it.
----------------------------------
Current markovify, existing chain JSON, state size 2:
14.08117773335427
Example: This is very easy.
----------------------------------
Current markovify, existing model JSON, state size 2:
15.376322960844263
Example: It may only be used on or associated in my hand in a note by the Underground and hurried from the room.
----------------------------------




Modified markovify, create new model, state size 2:
16.279055197359995
Example: He spoke in a few whispered words of his words, she gave a last hurried glance around, I saw the latter was your son.
----------------------------------
Modified markovify, existing chain JSON, state size 2:
13.715405708234757
Example: He calls attention, and with almost no restrictions whatsoever.
----------------------------------
Modified markovify, existing model JSON, state size 2:
6.346636804519221
Example: Instead of making up, and also in a note for me lay upon the previous night.
----------------------------------