chendl/compositional_test

1# Copyright 2023 The HuggingFace Team. All rights reserved.2#3# Licensed under the Apache License, Version 2.0 (the "License");4# you may not use this file except in compliance with the License.5# You may obtain a copy of the License at6#7#     http://www.apache.org/licenses/LICENSE-2.08#9# Unless required by applicable law or agreed to in writing, software10# distributed under the License is distributed on an "AS IS" BASIS,11# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.12# See the License for the specific language governing permissions and13# limitations under the License.14 15import unittest16 17from datasets import load_dataset18 19from transformers.pipelines import pipeline20from transformers.testing_utils import is_pipeline_test, nested_simplify, require_torch, slow21 22 23@is_pipeline_test24@require_torch25class ZeroShotAudioClassificationPipelineTests(unittest.TestCase):26    # Deactivating auto tests since we don't have a good MODEL_FOR_XX mapping,27    # and only CLAP would be there for now.28    # model_mapping = {CLAPConfig: CLAPModel}29 30    @require_torch31    def test_small_model_pt(self):32        audio_classifier = pipeline(33            task="zero-shot-audio-classification", model="hf-internal-testing/tiny-clap-htsat-unfused"34        )35        dataset = load_dataset("ashraq/esc50")36        audio = dataset["train"]["audio"][-1]["array"]37        output = audio_classifier(audio, candidate_labels=["Sound of a dog", "Sound of vaccum cleaner"])38        self.assertEqual(39            nested_simplify(output),40            [{"score": 0.501, "label": "Sound of a dog"}, {"score": 0.499, "label": "Sound of vaccum cleaner"}],41        )42 43    @unittest.skip("No models are available in TF")44    def test_small_model_tf(self):45        pass46 47    @slow48    @require_torch49    def test_large_model_pt(self):50        audio_classifier = pipeline(51            task="zero-shot-audio-classification",52            model="laion/clap-htsat-unfused",53        )54        # This is an audio of a dog55        dataset = load_dataset("ashraq/esc50")56        audio = dataset["train"]["audio"][-1]["array"]57        output = audio_classifier(audio, candidate_labels=["Sound of a dog", "Sound of vaccum cleaner"])58 59        self.assertEqual(60            nested_simplify(output),61            [62                {"score": 0.999, "label": "Sound of a dog"},63                {"score": 0.001, "label": "Sound of vaccum cleaner"},64            ],65        )66 67        output = audio_classifier([audio] * 5, candidate_labels=["Sound of a dog", "Sound of vaccum cleaner"])68        self.assertEqual(69            nested_simplify(output),70            [71                [72                    {"score": 0.999, "label": "Sound of a dog"},73                    {"score": 0.001, "label": "Sound of vaccum cleaner"},74                ],75            ]76            * 5,77        )78        output = audio_classifier(79            [audio] * 5, candidate_labels=["Sound of a dog", "Sound of vaccum cleaner"], batch_size=580        )81        self.assertEqual(82            nested_simplify(output),83            [84                [85                    {"score": 0.999, "label": "Sound of a dog"},86                    {"score": 0.001, "label": "Sound of vaccum cleaner"},87                ],88            ]89            * 5,90        )91 92    @unittest.skip("No models are available in TF")93    def test_large_model_tf(self):94        pass95