cffl/Exploring_Intelligent_Writing_Assistance

1# ###########################################################################2#3#  CLOUDERA APPLIED MACHINE LEARNING PROTOTYPE (AMP)4#  (C) Cloudera, Inc. 20225#  All rights reserved.6#7#  Applicable Open Source License: Apache 2.08#9#  NOTE: Cloudera open source products are modular software products10#  made up of hundreds of individual components, each of which was11#  individually copyrighted.  Each Cloudera open source product is a12#  collective work under U.S. Copyright Law. Your license to use the13#  collective work is as provided in your written agreement with14#  Cloudera.  Used apart from the collective work, this file is15#  licensed for your use pursuant to the open source license16#  identified above.17#18#  This code is provided to you pursuant a written agreement with19#  (i) Cloudera, Inc. or (ii) a third-party authorized to distribute20#  this code. If you do not have a written agreement with Cloudera nor21#  with an authorized and properly licensed third party, you do not22#  have any rights to access nor to use this code.23#24#  Absent a written agreement with Cloudera, Inc. (“Cloudera”) to the25#  contrary, A) CLOUDERA PROVIDES THIS CODE TO YOU WITHOUT WARRANTIES OF ANY26#  KIND; (B) CLOUDERA DISCLAIMS ANY AND ALL EXPRESS AND IMPLIED27#  WARRANTIES WITH RESPECT TO THIS CODE, INCLUDING BUT NOT LIMITED TO28#  IMPLIED WARRANTIES OF TITLE, NON-INFRINGEMENT, MERCHANTABILITY AND29#  FITNESS FOR A PARTICULAR PURPOSE; (C) CLOUDERA IS NOT LIABLE TO YOU,30#  AND WILL NOT DEFEND, INDEMNIFY, NOR HOLD YOU HARMLESS FOR ANY CLAIMS31#  ARISING FROM OR RELATED TO THE CODE; AND (D)WITH RESPECT TO YOUR EXERCISE32#  OF ANY RIGHTS GRANTED TO YOU FOR THE CODE, CLOUDERA IS NOT LIABLE FOR ANY33#  DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, PUNITIVE OR34#  CONSEQUENTIAL DAMAGES INCLUDING, BUT NOT LIMITED TO, DAMAGES35#  RELATED TO LOST REVENUE, LOST PROFITS, LOSS OF INCOME, LOSS OF36#  BUSINESS ADVANTAGE OR UNAVAILABILITY, OR LOSS OR CORRUPTION OF37#  DATA.38#39# ###########################################################################40 41import pytest42import transformers43 44from src.style_transfer import StyleTransfer45from src.style_classification import StyleIntensityClassifier46from src.content_preservation import ContentPreservationScorer47from src.transformer_interpretability import InterpretTransformer48 49 50@pytest.fixture51def subjectivity_example_data():52    examples = [53        """there is an iconic roadhouse, named "spud's roadhouse", which sells fuel and general shop items , has great meals and has accommodation.""",54        "chemical abstracts service (cas), a prominent division of the american chemical society, is the world's leading source of chemical information.",55        "the most serious scandal was the iran-contra affair.",56        "another strikingly elegant four-door saloon for the s3 continental came from james young.",57        "other ambassadors also sent their messages of condolence following her passing.",58    ]59 60    ground_truth = [61        'there is a roadhouse, named "spud\'s roadhouse", which sells fuel and general shop items and has accommodation.',62        "chemical abstracts service (cas), a division of the american chemical society, is a source of chemical information.",63        "one controversy was the iran-contra affair.",64        "another four-door saloon for the s3 continental came from james young.",65        "other ambassadors also sent their messages of condolence following her death.",66    ]67 68    return {"examples": examples, "ground_truth": ground_truth}69 70 71@pytest.fixture72def subjectivity_styletransfer():73    MODEL_PATH = "cffl/bart-base-styletransfer-subjective-to-neutral"74    return StyleTransfer(model_identifier=MODEL_PATH, max_gen_length=200)75 76 77@pytest.fixture78def subjectivity_styleintensityclassifier():79    CLS_MODEL_PATH = "cffl/bert-base-styleclassification-subjective-neutral"80    return StyleIntensityClassifier(model_identifier=CLS_MODEL_PATH)81 82 83@pytest.fixture84def subjectivity_contentpreservationscorer():85    CLS_MODEL_PATH = "cffl/bert-base-styleclassification-subjective-neutral"86    SBERT_MODEL_PATH = "sentence-transformers/all-MiniLM-L6-v2"87    return ContentPreservationScorer(88        cls_model_identifier=CLS_MODEL_PATH, sbert_model_identifier=SBERT_MODEL_PATH89    )90 91 92@pytest.fixture93def subjectivity_interprettransformer():94    CLS_MODEL_PATH = "cffl/bert-base-styleclassification-subjective-neutral"95    return InterpretTransformer(cls_model_identifier=CLS_MODEL_PATH)96 97 98# test class initialization99def test_StyleTransfer_init(subjectivity_styletransfer):100    assert isinstance(101        subjectivity_styletransfer.pipeline,102        transformers.pipelines.text2text_generation.Text2TextGenerationPipeline,103    )104 105 106def test_StyleIntensityClassifier_init(subjectivity_styleintensityclassifier):107    assert isinstance(108        subjectivity_styleintensityclassifier.pipeline,109        transformers.pipelines.text_classification.TextClassificationPipeline,110    )111 112 113def test_ContentPreservationScorer_init(subjectivity_contentpreservationscorer):114    assert isinstance(115        subjectivity_contentpreservationscorer.cls_model,116        transformers.models.bert.modeling_bert.BertForSequenceClassification,117    )118    assert isinstance(119        subjectivity_contentpreservationscorer.sbert_model,120        transformers.models.bert.modeling_bert.BertModel,121    )122 123 124def test_InterpretTransformer_init(subjectivity_interprettransformer):125    assert isinstance(126        subjectivity_interprettransformer.cls_model,127        transformers.models.bert.modeling_bert.BertForSequenceClassification,128    )129 130 131# test class functionality132def test_StyleTransfer_transfer(subjectivity_styletransfer, subjectivity_example_data):133    assert subjectivity_example_data[134        "ground_truth"135    ] == subjectivity_styletransfer.transfer(subjectivity_example_data["examples"])136 137 138def test_StyleIntensityClassifier_calculate_transfer_intensity_fraction(139    subjectivity_styleintensityclassifier, subjectivity_example_data140):141    sti_frac = (142        subjectivity_styleintensityclassifier.calculate_transfer_intensity_fraction(143            input_text=subjectivity_example_data["examples"],144            output_text=subjectivity_example_data["ground_truth"],145        )146    )147    assert sti_frac == [148        0.9891820847234861,149        0.9808499743983614,150        0.8070009460737938,151        0.9913705583756346,152        0.9611679711017459,153    ]154 155 156def test_ContentPreservationScorer_calculate_content_preservation_score(157    subjectivity_contentpreservationscorer, subjectivity_example_data158):159    cps = subjectivity_contentpreservationscorer.calculate_content_preservation_score(160        input_text=subjectivity_example_data["examples"],161        output_text=subjectivity_example_data["ground_truth"],162        mask_type="none",163    )164    assert cps == [0.9369, 0.9856, 0.7328, 0.9718, 0.9709]165