config.py

# coding=utf-8
# !/usr/bin/python3.6  # Please use python 3.6
"""
__synopsis__    : Contains system configuration details and other global variables.

__description__ : Benefit: We can print the configuration on every run and get the hyper-parameter configuration for each run.
__project__     : MNXC
__author__      : Samujjwal Ghosh <cs16resch01001@iith.ac.in>
__version__     : ": 0.1 "
__date__        : "04-03-2019"
__copyright__   : "Copyright (c) 2019, All rights reserved."
__license__     : "This source code is licensed under the MIT-style license found in the LICENSE file in the root
                  directory of this source tree."

__classes__     : config

__variables__   : configuration, seed, platform

__methods__     :

__last_modified__:
"""

import json
from logger.logger import logger

global seed
seed = 0

global configuration
configuration = {
    "data"        :{
        "dataset_name":"Wiki10-31K_onehot",
        "val_split"   :0.1,
        "test_split"  :0.3,
        "show_stat"   :False
    },

    "xc_datasets" :{
        "Wiki10-31K"        :"html",
        "Wiki10-31K_onehot" :"html",
        "Wiki10-31K_fixed5" :"html",
        "Wiki10-31K_pointer":"html",
        "AmazonCat-14K"     :"json",
        "Amazon-3M"         :"json",
        "Delicious-T140"    :"html",
        "AmazonCat-13K"     :"txt",
        "Amazon-670K"       :"txt",
        "EurLex"            :"uniq"
    },

    "model"       :{
        "num_folds"           :5,
        "max_sequence_length" :100,
        "max_vec_len"         :5000,
        "dropout"             :0.1,
        "dropout_external"    :0.0,
        "clipnorm"            :1.0,
        "data_slice"          :5120,

        "g_encoder"           :"cnn",
        "use_cuda"            :False,
        "normalize_inputs"    :False,
        "tfidf_avg"           :False,
        "kernel_size"         :1,
        "stride"              :1,
        "padding"             :1,
        "context"             :10,
        "classify_count"      :0,
        "fce"                 :True,
        "optimizer"           :{
            "optimizer_type":"adam",
            "learning_rate" :3e-4,
            "lr_decay"      :0,
            "weight_decay"  :0,
            "momentum"      :0.9,
            "dampening"     :0.9,
            "alpha"         :0.99,
            "rho"           :0.9,
            "centered"      :False
        },
        "view_grads"          :False,
        "view_train_precision":True
    },

    "lstm_params" :{
        "num_layers"   :1,
        "bias"         :True,
        "batch_first"  :True,
        "bidirectional":True,
        "hid_size"     :3,
    },

    "cnn_params"  :{
        "padding"    :1,
        "stride"     :1,
        "kernel_size":1,
        "bias"       :True,
    },

    "sampling"    :{
        "num_epochs"           :20,
        "num_train_epoch"      :5,
        "batch_size"           :2,
        "categories_per_batch" :2,
        "supports_per_category":2,
        "targets_per_category" :2
    },

    "prep_vecs"   :{
        "max_nb_words"      :20000,
        "min_word_count"    :1,
        "window"            :7,
        "min_count"         :1,
        "negative"          :10,
        "num_chunks"        :10,
        "vectorizer"        :"doc2vec",
        "sample_repeat_mode":"append",
        "input_size"        :7,
        "tfidf_avg"         :False,
        "idf"               :True
    },

    "text_process":{
        "encoding"        :'latin-1',
        "sents_chunk_mode":"word_avg",
        "workers"         :5
    },

    "paths"       :{
        "result_file" :"result.txt",
        "log_dir"     :"/logs",

        "pretrain_dir":{
            "Windows":"D:\\Datasets\\Extreme Classification",
            "OSX"    :"/home/cs16resch01001/datasets/Extreme Classification",
            "Linux"  :{
                "sam"           :"/media/sam/Data/Datasets/pretrain",  ## Code path: /home/cs14resch11001/codes/MNXC
                "cs14resch11001":"/raid/ravi/pretrain"
            }
        },

        "dataset_dir" :{
            "Windows":"D:\\Datasets\\Extreme Classification",
            "OSX"    :"/home/cs16resch01001/datasets/Extreme Classification",
            "Linux"  :{
                "sam"           :"/media/sam/Data/Datasets/Extreme Classification",
                "cs14resch11001":"/raid/ravi/Datasets/Extreme Classification"
            }
        }
    },

    "html_parser" :{
        "alt_text"              :True,
        "ignore_table"          :True,
        "decode_errors"         :"ignore",
        "default_alt"           :"<IMG>",
        "ignore_link"           :True,
        "reference_links"       :True,
        "bypass_tables"         :True,
        "ignore_emphasis"       :True,
        "unicode_snob"          :True,
        "no_automatic_links"    :True,
        "no_skip_internal_links":True,
        "single_line_break"     :True,
        "escape_all"            :True
    }
}


class Config(object):
    """ Contains all configuration details of the project. """

    def __init__(self):
        super(Config,self).__init__()

        self.configuration = configuration

    def get_config(self):
        """

        :return:
        """
        return self.configuration

    def print_config(self,indent=4,sort=True):
        """ Prints the config. """
        logger.info("[{}] : {}".format("Configuration",json.dumps(self.configuration,indent=indent,sort_keys=sort)))

    @staticmethod
    def get_platform():
        """ Returns dataset path based on OS.

        :return: str
        """
        import platform

        if platform.system() == 'Windows':
            return platform.system()
        elif platform.system() == 'Linux':
            return platform.system()
        else:  ## OS X returns name 'Darwin'
            return "OSX"

    @staticmethod
    def get_username():
        """
        :returns the current username.

        :return: string
        """
        try:
            import os,pwd

            username = pwd.getpwuid(os.getuid()).pw_name
        except Exception as e:
            import getpass

            username = getpass.getuser()
        # finally:
        #     username = os.environ.get('USER')

        return username


config_cls = Config()
config_cls.print_config()

global platform
platform = config_cls.get_platform()
global username
username = config_cls.get_username()


def main():
    """
    Main module to start code
    :param args:
        Type: tuple
        Required
        Read Only
    :return:
    """
    pass


if __name__ == "__main__":
    main()