change directory name

### Data Preparation

1. Dataset Download: [Wizard of Wikipedia](https://parl.ai/projects/wizard_of_wikipedia/) and [Wizard of Internet](https://parl.ai/projects/sea/)

2. Data Processing: We provide the script to run the [`data processing`](../../examples/knwl_dialo/data_processing.sh) of the datatsets.

2. Data Processing: We provide the script to run the [`data processing`](../../examples/msdp/data_processing.sh) of the datatsets.

### Stage-1: Prompting for Knowledge Generation

1. We provide the script to perform the [`first-stage prompting`](../../examples/knwl_dialo/prompt_knwl_gen.sh) for the knowledge generation.

2. We provide the [`evaluation script`](../../examples/knwl_dialo/eval_knwl_generation.sh) for the automatic evaluation (i.e., F1, BLEU, METEOR, and ROUGE-L) of the knowledge generation.

1. We provide the script to perform the [`first-stage prompting`](../../examples/msdp/prompt_knwl_gen.sh) for the knowledge generation.

2. We provide the [`evaluation script`](../../examples/msdp/eval_knwl_generation.sh) for the automatic evaluation (i.e., F1, BLEU, METEOR, and ROUGE-L) of the knowledge generation.

### Stage-2: Prompting for Response Generation

1. We provide the script to [`prepare the input file`](../../examples/knwl_dialo/prep_resp_gen.sh) for the response generation (based on the previously generated knowledge file).

2. We provide the script to perform the [`second-stage prompting`](../../examples/knwl_dialo/prompt_resp_gen.sh) for the response generation.

3.  We provide the [`evaluation script`](../../examples/knwl_dialo/eval_resp_generation.sh) for the automatic evaluation (i.e., F1, KF1, BLEU, METEOR, and ROUGE-L) of the response generation.

1. We provide the script to [`prepare the input file`](../../examples/msdp/prep_resp_gen.sh) for the response generation (based on the previously generated knowledge file).

2. We provide the script to perform the [`second-stage prompting`](../../examples/msdp/prompt_resp_gen.sh) for the response generation.

3.  We provide the [`evaluation script`](../../examples/msdp/eval_resp_generation.sh) for the automatic evaluation (i.e., F1, KF1, BLEU, METEOR, and ROUGE-L) of the response generation.

# coding=utf-8

# Copyright (c) 2020, NVIDIA CORPORATION.  All rights reserved.

#

# Licensed under the Apache License, Version 2.0 (the "License");

# you may not use this file except in compliance with the License.

# You may obtain a copy of the License at

#

#     http://www.apache.org/licenses/LICENSE-2.0

#

# Unless required by applicable law or agreed to in writing, software

# distributed under the License is distributed on an "AS IS" BASIS,

# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.

# See the License for the specific language governing permissions and

# limitations under the License.

"""Model evaluation"""

from megatron import get_args

from megatron import print_rank_0

from tasks.msdp.metrics import F1Metric

from tqdm import tqdm

def evaluate_f1(guess_file, answer_file):

    """Evaluating F1 Score"""

    guess_list = []

    print_rank_0('reading %s' % guess_file)

    with open(guess_file, "r") as f:

        for i, line in enumerate(tqdm(f)):

            line = line.strip()

            if "<|endoftext|>" in line:

                line = line.replace("<|endoftext|>", "")

            guess_list.append(line)

    answer_list = []

    print_rank_0('reading %s' % answer_file)

    with open(answer_file, "r") as f:

        for i, line in enumerate(tqdm(f)):

            line = line.strip()

            if line == "no_passages_used":

                line = ""

            answer_list.append(line)

    assert len(guess_list) == len(answer_list), \

        "lengths of guess and answer are different!"

    precision, recall, f1 = F1Metric.compute_all_pairs(guess_list, answer_list)

    print_rank_0('Precision: %.4f; recall: %.4f; f1: %.4f' % (precision, recall, f1))

    print_rank_0('done :-)')

def main():

    args = get_args()

    evaluate_f1(args.guess_file, args.answer_file)

# coding=utf-8

# Copyright (c) 2020, NVIDIA CORPORATION.  All rights reserved.

#

# Licensed under the Apache License, Version 2.0 (the "License");

# you may not use this file except in compliance with the License.

# You may obtain a copy of the License at

#

#     http://www.apache.org/licenses/LICENSE-2.0

#

# Unless required by applicable law or agreed to in writing, software

# distributed under the License is distributed on an "AS IS" BASIS,

# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.

# See the License for the specific language governing permissions and

# limitations under the License.

"""Run multi-stage dialogue prompting (MSDP)."""

import os

import sys

sys.path.append(os.path.abspath(os.path.join(

    os.path.join(os.path.dirname(__file__), os.path.pardir), os.path.pardir)))

from megatron import get_args

from megatron.initialize import initialize_megatron

def get_tasks_args(parser):

    """Provide extra arguments required for tasks."""

    group = parser.add_argument_group(title='tasks')

    # parameters for the knowledgeable dialogue generation

    group.add_argument('--task', type=str, required=True,

                       help='Task name.')

    group.add_argument("--sample-input-file", type=str, default=None,

                       help='Get input from file instead of interactive mode, '

                       'each line is an input.')

    group.add_argument("--sample-output-file", type=str, default=None,

                       help='Output file got from --sample-input-file')

    group.add_argument('--prompt-file', type=str, default=None,

                       help='prompting file')

    group.add_argument('--prompt-type', type=str, default=None, 

                       choices=['knowledge', 'response'],

                       help='prompt type (knowledge or response)')

    group.add_argument('--num-prompt-examples', type=int, default=10,

                       help='number of prompt examples')

    group.add_argument('--guess-file', type=str, default=None,

                       help='datapath for generated sentences')

    group.add_argument('--answer-file', type=str, default=None,

                       help='datapath for golden sentences')

    group.add_argument('--out-seq-length', type=int, default=100,

                       help='output sequence length')

    group.add_argument('--api-prompt', default=False, action="store_true",

                       help='setup model api for prompting')

    group.add_argument('--megatron-api-url', type=str, default=None,

                       help='url of the megatron api')

    return parser

if __name__ == '__main__':

    initialize_megatron(extra_args_provider=get_tasks_args)

    args = get_args()

    if args.num_layers_per_virtual_pipeline_stage is not None:

        print("Interleaved pipeline schedule is not yet supported for downstream tasks.")

        exit()

    if args.task == 'MSDP-PROMPT':

        from tasks.msdp.prompt import main

    elif args.task == 'MSDP-EVAL-F1':

        from tasks.msdp.evaluate import main

    else:

        raise NotImplementedError('Task {} is not implemented.'.format(

            args.task))

    main()

# The following code is adapted from

# https://github.com/facebookresearch/ParlAI/blob/master/parlai/core/metrics.py, 

# which is licensed under the MIT license. More details on the license can be 

# found at https://github.com/facebookresearch/ParlAI/blob/master/LICENSE.

"""Provides standard metric evaluations for dialog."""

from collections import Counter

from typing import List

import numpy as np

import re

re_art = re.compile(r'\b(a|an|the)\b')

re_punc = re.compile(r'[!"#$%&()*+,-./:;<=>?@\[\]\\^`{|}~_\']')

def normalize_answer(s):

    """

    Lower text and remove punctuation, articles and extra whitespace.

    """

    s = s.lower()

    s = re_punc.sub(' ', s)

    s = re_art.sub(' ', s)

    s = ' '.join(s.split())

    return s

class F1Metric:

    """

    Helper class which computes token-level F1.

    """

    @staticmethod

    def _prec_recall_f1_score(pred_items, gold_items):

        """

        Compute precision, recall and f1 given a set of gold and prediction items.

        :param pred_items: iterable of predicted values

        :param gold_items: iterable of gold values

        :return: tuple (p, r, f1) for precision, recall, f1

        """

        common = Counter(gold_items) & Counter(pred_items)

        num_same = sum(common.values())

        if num_same == 0:

            return 0, 0, 0

        precision = 1.0 * num_same / len(pred_items)

        recall = 1.0 * num_same / len(gold_items)

        f1 = (2 * precision * recall) / (precision + recall)

        return precision, recall, f1

    @staticmethod

    def compute_each_pair(guess: str, answer: str):

        if answer == "":

            return None, None, None

        if guess == "":

            return 0, 0, 0

        g_tokens = normalize_answer(guess).split()

        a_tokens = normalize_answer(answer).split()

        precision, recall, f1 = F1Metric._prec_recall_f1_score(g_tokens, a_tokens)

        return precision, recall, f1

    @staticmethod

    def compute_all_pairs(guesses: List[str], answers: List[str]):

        # additional augment:

        assert len(guesses) == len(answers)

        precision_list, recall_list, f1_list = [], [], []

        for guess, answer in zip(guesses, answers):

            precision, recall, f1 = F1Metric.compute_each_pair(guess, answer)

            if precision is None or recall is None or f1 is None:

                continue

            precision_list.append(precision)

            recall_list.append(recall)

            f1_list.append(f1)

        return np.mean(precision_list), np.mean(recall_list), np.mean(f1_list)