merge with master

.dev_scripts/run_docker.sh

@@ -0,0 +1,7 @@
+#sudo docker run --name zwm_maas -v /home/wenmeng.zwm/workspace:/home/wenmeng.zwm/workspace   --net host -ti reg.docker.alibaba-inc.com/pai-dlc/tensorflow-training:2.3-gpu-py36-cu101-ubuntu18.04 bash
+#sudo docker run --name zwm_maas_pytorch -v /home/wenmeng.zwm/workspace:/home/wenmeng.zwm/workspace --net host  -ti reg.docker.alibaba-inc.com/pai-dlc/pytorch-training:1.10PAI-gpu-py36-cu113-ubuntu18.04 bash
+CONTAINER_NAME=modelscope-dev
+IMAGE_NAME=registry.cn-shanghai.aliyuncs.com/modelscope/modelscope
+IMAGE_VERSION=v0.1.1-16-g62856fa-devel
+MOUNT_DIR=/home/wenmeng.zwm/workspace
+sudo docker run --name  $CONTAINER_NAME -v $MOUNT_DIR:$MOUNT_DIR --net host  -ti ${IMAGE_NAME}:${IMAGE_VERSION} bash

.gitattributes

@@ -1,3 +1,4 @@
 *.png filter=lfs diff=lfs merge=lfs -text
 *.jpg filter=lfs diff=lfs merge=lfs -text
 *.mp4 filter=lfs diff=lfs merge=lfs -text
+*.wav filter=lfs diff=lfs merge=lfs -text

.gitignore

@@ -124,7 +124,3 @@ replace.sh
 
 # Pytorch
 *.pth
-
-
-# audio
-*.wav

data/test/audios/kws_bofangyinyue.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a72a7b8d1e8be6ebaa09aeee0d71472569bc62cc4872ecfdbd1651bb3d03eaba
+size 69110

data/test/audios/kws_xiaoyunxiaoyun.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c6b1671bcfa872278c99490cd1acb08297b8df4dc78f268e4b6a582b4364e4a1
+size 297684

docker/pytorch.dockerfile

@@ -30,7 +30,8 @@ RUN apt-get update &&\
     zip \
     zlib1g-dev \
     unzip \
-    pkg-config
+    pkg-config \
+    libsndfile1
 
 # install modelscope and its python env
 WORKDIR /opt/modelscope

docs/source/index.rst

@@ -13,6 +13,7 @@ ModelScope doc
 
    quick_start.md
    develop.md
+   faq.md
 
 .. toctree::
    :maxdepth: 2
@@ -20,6 +21,8 @@ ModelScope doc
 
    tutorials/index
 
+
+
 .. toctree::
    :maxdepth: 2
    :caption: Changelog

docs/source/quick_start.md

@@ -1,72 +1,55 @@
 # 快速开始
-
+ModelScope Library目前支持tensorflow，pytorch深度学习框架进行模型训练、推理， 在Python 3.7+, Pytorch 1.8+, Tensorflow1.15+，Tensorflow 2.6上测试可运行。
+注： 当前（630）版本仅支持python3.7 以及linux环境，其他环境(mac,windows等)支持预计730完成。
 ## python环境配置
 首先，参考[文档](https://docs.anaconda.com/anaconda/install/) 安装配置Anaconda环境
 
 安装完成后，执行如下命令为modelscope library创建对应的python环境。
 ```shell
-conda create -n modelscope python=3.6
+conda create -n modelscope python=3.7
 conda activate modelscope
 ```
-检查python和pip命令是否切换到conda环境下。
+## 安装深度学习框架
+* 安装pytorch[参考链接](https://pytorch.org/get-started/locally/)
 ```shell
-which python
-# ~/workspace/anaconda3/envs/modelscope/bin/python
-
-which pip
-# ~/workspace/anaconda3/envs/modelscope/bin/pip
+pip install torch torchvision
 ```
-注： 本项目只支持`python3`环境，请勿使用python2环境。
-
-## 第三方依赖安装
-
-ModelScope Library目前支持tensorflow，pytorch两大深度学习框架进行模型训练、推理， 在Python 3.6+,  Pytorch 1.8+, Tensorflow 2.6上测试可运行，用户可以根据所选模型对应的计算框架进行安装，可以参考如下链接进行安装所需框架:
-
-* [Pytorch安装指导](https://pytorch.org/get-started/locally/)
-* [Tensorflow安装指导](https://www.tensorflow.org/install/pip)
-
-部分第三方依赖库需要提前安装numpy
-```
-pip install numpy
+* 安装Tensorflow[参考链接](https://www.tensorflow.org/install/pip)
+```shell
+pip install --upgrade tensorflow
 ```
-
 ## ModelScope library 安装
 
 注： 如果在安装过程中遇到错误，请前往[常见问题](faq.md)查找解决方案。
 
 ### pip安装
+执行如下命令：
 ```shell
-pip install -r http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/release/maas/modelscope.txt
+pip install model_scope[all] -f https://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/release/maas/repo.html
 ```
-
-安装成功后，可以执行如下命令进行验证安装是否正确
-```shell
-python -c "from modelscope.pipelines import pipeline;print(pipeline('image-matting',model='damo/image-matting-person')('http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/data/test/maas/image_matting/test.png'))"
-```
-
-
 ### 使用源码安装
-
 适合本地开发调试使用，修改源码后可以直接执行
+下载源码可以直接clone代码到本地
 ```shell
 git clone git@gitlab.alibaba-inc.com:Ali-MaaS/MaaS-lib.git modelscope
 git fetch origin master
 git checkout master
-
 cd modelscope
-
-#安装依赖
+```
+安装依赖并设置PYTHONPATH
+```shell
 pip install -r requirements.txt
-
-# 设置PYTHONPATH
 export PYTHONPATH=`pwd`
 ```
-
+### 安装验证
 安装成功后，可以执行如下命令进行验证安装是否正确
 ```shell
-python -c "from modelscope.pipelines import pipeline;print(pipeline('image-matting',model='damo/image-matting-person')('http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/data/test/maas/image_matting/test.png'))"
+python -c "from modelscope.pipelines import pipeline;print(pipeline('word-segmentation')('今天天气不错，适合 出去游玩'))"
+{'output': '今天 天气 不错 ， 适合 出去 游玩'}
 ```
+## 推理
 
+pipeline函数提供了简洁的推理接口，相关介绍和示例请参考[pipeline使用教程](tutorials/pipeline.md)
 
 ## 训练
 
@@ -75,46 +58,3 @@ to be done
 ## 评估
 
 to be done
-
-## 推理
-
-pipeline函数提供了简洁的推理接口，示例如下， 更多pipeline介绍和示例请参考[pipeline使用教程](tutorials/pipeline.md)
-
-```python
-import cv2
-import os.path as osp
-from modelscope.pipelines import pipeline
-from modelscope.utils.constant import Tasks
-
-# 根据任务名创建pipeline
-img_matting = pipeline(Tasks.image_matting, model='damo/image-matting-person')
-
-# 直接提供图像文件的url作为pipeline推理的输入
-result = img_matting(
-    'http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/data/test/maas/image_matting/test.png'
-)
-cv2.imwrite('result.png', result['output_png'])
-print(f'Output written to {osp.abspath("result.png")}')
-
-```
-
-此外，pipeline接口也能接收Dataset作为输入，上面的代码同样可以实现为
-
-```python
-import cv2
-import os.path as osp
-from modelscope.pipelines import pipeline
-from modelscope.utils.constant import Tasks
-from modelscope.msdatasets import MsDataset
-
-# 使用图像url构建MsDataset，此处也可通过 input_location = '/dir/to/images' 来使用本地文件夹
-input_location = [
-    'http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/data/test/maas/image_matting/test.png'
-]
-dataset = MsDataset.load(input_location, target='image')
-img_matting = pipeline(Tasks.image_matting, model='damo/image-matting-person')
-# 输入为MsDataset时，输出的结果为迭代器
-result = img_matting(dataset)
-cv2.imwrite('result.png', next(result)['output_png'])
-print(f'Output written to {osp.abspath("result.png")}')
-```

docs/source/tutorials/pipeline.md

@@ -1,84 +1,62 @@
 # Pipeline使用教程
-
-本文将简单介绍如何使用`pipeline`函数加载模型进行推理。`pipeline`函数支持按照任务类型、模型名称从模型仓库
-拉取模型进行进行推理，当前支持的任务有
-
-* 人像抠图 (image-matting)
-* 基于bert的语义情感分析 (bert-sentiment-analysis)
-
-本文将从如下方面进行讲解如何使用Pipeline模块：
+本文简单介绍如何使用`pipeline`函数加载模型进行推理。`pipeline`函数支持按照任务类型、模型名称从模型仓库拉取模型进行进行推理，包含以下几个方面：
 * 使用pipeline()函数进行推理
 * 指定特定预处理、特定模型进行推理
 * 不同场景推理任务示例
-
 ## 环境准备
 详细步骤可以参考 [快速开始](../quick_start.md)
-
 ## Pipeline基本用法
+下面以中文分词任务为例，说明pipeline函数的基本用法
 
-1. pipeline函数支持指定特定任务名称，加载任务默认模型，创建对应Pipeline对象
+1. pipeline函数支持指定特定任务名称，加载任务默认模型，创建对应pipeline对象
    执行如下python代码
    ```python
-   >>> from modelscope.pipelines import pipeline
-   >>> img_matting = pipeline(task='image-matting', model='damo/image-matting-person')
+   from modelscope.pipelines import pipeline
+   word_segmentation = pipeline('word-segmentation')
    ```
 
-2. 传入单张图像url进行处理
+2. 输入文本
    ``` python
-   >>> import cv2
-   >>> result = img_matting('http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/data/test/maas/image_matting/test.png')
-   >>> cv2.imwrite('result.png', result['output_png'])
-   >>> import os.path as osp
-   >>> print(f'result file path is {osp.abspath("result.png")}')
+   input = '今天天气不错，适合出去游玩'
+   print(word_segmentation(input))
+   {'output': '今天 天气 不错 ， 适合 出去 游玩'}
    ```
 
-   pipeline对象也支持传入一个列表输入，返回对应输出列表，每个元素对应输入样本的返回结果
-   ```python
-   >>> results = img_matting(
-       [
-           'http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/data/test/maas/image_matting/test.png',
-           'http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/data/test/maas/image_matting/test.png',
-           'http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/data/test/maas/image_matting/test.png',
-       ])
-   ```
+3. 输入多条样本
+
+pipeline对象也支持传入多个样本列表输入，返回对应输出列表，每个元素对应输入样本的返回结果
 
-   如果pipeline对应有一些后处理参数，也支持通过调用时候传入.
    ```python
-   >>> pipe = pipeline(task_name)
-   >>> result = pipe(input, post_process_args)
+   inputs =  ['今天天气不错，适合出去游玩','这本书很好，建议你看看']
+   print(word_segmentation(inputs))
+   [{'output': '今天 天气 不错 ， 适合 出去 游玩'}, {'output': '这 本 书 很 好 ， 建议 你 看看'}]
    ```
-
 ## 指定预处理、模型进行推理
 pipeline函数支持传入实例化的预处理对象、模型对象，从而支持用户在推理过程中定制化预处理、模型。
-下面以文本情感分类为例进行介绍。
 
-由于demo模型为EasyNLP提供的模型，首先，安装EasyNLP
-```shell
-pip install https://atp-modelzoo-sh.oss-cn-shanghai.aliyuncs.com/release/package/whl/easynlp-0.0.4-py2.py3-none-any.whl
-```
-
-
-下载模型文件
-```shell
-wget https://atp-modelzoo-sh.oss-cn-shanghai.aliyuncs.com/release/easynlp_modelzoo/alibaba-pai/bert-base-sst2.zip && unzip bert-base-sst2.zip
-```
-
-创建tokenizer和模型
+1. 首先，创建预处理方法和模型
 ```python
->>> from modelscope.models import Model
->>> from modelscope.preprocessors import SequenceClassificationPreprocessor
->>> model = Model.from_pretrained('damo/bert-base-sst2')
->>> tokenizer = SequenceClassificationPreprocessor(
-            model.model_dir, first_sequence='sentence', second_sequence=None)
+from modelscope.models import Model
+from modelscope.preprocessors import TokenClassifcationPreprocessor
+model = Model.from_pretrained('damo/nlp_structbert_word-segmentation_chinese-base')
+tokenizer = TokenClassifcationPreprocessor(model.model_dir)
 ```
 
-使用tokenizer和模型对象创建pipeline
+2. 使用tokenizer和模型对象创建pipeline
 ```python
->>> from modelscope.pipelines import pipeline
->>> semantic_cls = pipeline('text-classification', model=model,   preprocessor=tokenizer)
->>> semantic_cls("Hello world!")
+from modelscope.pipelines import pipeline
+word_seg = pipeline('word-segmentation', model=model, preprocessor=tokenizer)
+input = '今天天气不错，适合出去游玩'
+print(word_seg(input))
+{'output': '今天 天气 不错 ， 适合 出去 游玩'}
 ```
-
 ## 不同场景任务推理示例
-
-人像抠图、语义分类建上述两个例子。  其他例子未来添加。
+下面以一个图像任务：人像抠图（'image-matting'）为例，进一步说明pipeline的用法
+```python
+import cv2
+import os.path as osp
+from modelscope.pipelines import pipeline
+img_matting = pipeline('image-matting')
+result = img_matting('http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/data/test/maas/image_matting/test.png')
+cv2.imwrite('result.png', result['output_png'])
+```

modelscope/hub/api.py

@@ -1,6 +1,8 @@
 import os
 import pickle
+import shutil
 import subprocess
+from collections import defaultdict
 from http.cookiejar import CookieJar
 from os.path import expanduser
 from typing import List, Optional, Tuple, Union
@@ -8,8 +10,11 @@ from typing import List, Optional, Tuple, Union
 import requests
 
 from modelscope.utils.logger import get_logger
+from ..msdatasets.config import DOWNLOADED_DATASETS_PATH, HUB_DATASET_ENDPOINT
+from ..utils.constant import DownloadMode
 from .constants import MODELSCOPE_URL_SCHEME
-from .errors import InvalidParameter, NotExistError, is_ok, raise_on_error
+from .errors import (InvalidParameter, NotExistError, datahub_raise_on_error,
+                     is_ok, raise_on_error)
 from .utils.utils import (get_endpoint, get_gitlab_domain,
                           model_id_to_group_owner_name)
 
@@ -18,8 +23,9 @@ logger = get_logger()
 
 class HubApi:
 
-    def __init__(self, endpoint=None):
+    def __init__(self, endpoint=None, dataset_endpoint=None):
         self.endpoint = endpoint if endpoint is not None else get_endpoint()
+        self.dataset_endpoint = dataset_endpoint if dataset_endpoint is not None else HUB_DATASET_ENDPOINT
 
     def login(
         self,
@@ -241,6 +247,70 @@ class HubApi:
             files.append(file)
         return files
 
+    def list_datasets(self):
+        path = f'{self.dataset_endpoint}/api/v1/datasets'
+        headers = None
+        params = {}
+        r = requests.get(path, params=params, headers=headers)
+        r.raise_for_status()
+        dataset_list = r.json()['Data']
+        return [x['Name'] for x in dataset_list]
+
+    def fetch_dataset_scripts(self,
+                              dataset_name: str,
+                              namespace: str,
+                              download_mode: Optional[DownloadMode],
+                              version: Optional[str] = 'master'):
+        if namespace is None:
+            raise ValueError(
+                f'Dataset from Hubs.modelscope should have a valid "namespace", but get {namespace}'
+            )
+        version = version or 'master'
+        cache_dir = os.path.join(DOWNLOADED_DATASETS_PATH, dataset_name,
+                                 namespace, version)
+        download_mode = DownloadMode(download_mode
+                                     or DownloadMode.REUSE_DATASET_IF_EXISTS)
+        if download_mode == DownloadMode.FORCE_REDOWNLOAD and os.path.exists(
+                cache_dir):
+            shutil.rmtree(cache_dir)
+        os.makedirs(cache_dir, exist_ok=True)
+        datahub_url = f'{self.dataset_endpoint}/api/v1/datasets/{namespace}/{dataset_name}'
+        r = requests.get(datahub_url)
+        resp = r.json()
+        datahub_raise_on_error(datahub_url, resp)
+        dataset_id = resp['Data']['Id']
+        datahub_url = f'{self.dataset_endpoint}/api/v1/datasets/{dataset_id}/repo/tree?Revision={version}'
+        r = requests.get(datahub_url)
+        resp = r.json()
+        datahub_raise_on_error(datahub_url, resp)
+        file_list = resp['Data']
+        if file_list is None:
+            raise NotExistError(
+                f'The modelscope dataset [dataset_name = {dataset_name}, namespace = {namespace}, '
+                f'version = {version}] dose not exist')
+
+        file_list = file_list['Files']
+        local_paths = defaultdict(list)
+        for file_info in file_list:
+            file_path = file_info['Path']
+            if file_path.endswith('.py'):
+                datahub_url = f'{self.dataset_endpoint}/api/v1/datasets/{dataset_id}/repo/files?' \
+                              f'Revision={version}&Path={file_path}'
+                r = requests.get(datahub_url)
+                r.raise_for_status()
+                content = r.json()['Data']['Content']
+                local_path = os.path.join(cache_dir, file_path)
+                if os.path.exists(local_path):
+                    logger.warning(
+                        f"Reusing dataset {dataset_name}'s python file ({local_path})"
+                    )
+                    local_paths['py'].append(local_path)
+                    continue
+                with open(local_path, 'w') as f:
+                    f.writelines(content)
+                local_paths['py'].append(local_path)
+        return local_paths
+
 
 class ModelScopeConfig:
     path_credential = expanduser('~/.modelscope/credentials')

modelscope/hub/constants.py

@@ -1,6 +1,8 @@
 MODELSCOPE_URL_SCHEME = 'http://'
-DEFAULT_MODELSCOPE_DOMAIN = '47.94.223.21:31090'
+DEFAULT_MODELSCOPE_IP = '47.94.223.21'
+DEFAULT_MODELSCOPE_DOMAIN = DEFAULT_MODELSCOPE_IP + ':31090'
 DEFAULT_MODELSCOPE_GITLAB_DOMAIN = '101.201.119.157:31102'
+DEFAULT_MODELSCOPE_DATA_ENDPOINT = MODELSCOPE_URL_SCHEME + DEFAULT_MODELSCOPE_IP + ':31752'
 
 DEFAULT_MODELSCOPE_GROUP = 'damo'
 MODEL_ID_SEPARATOR = '/'

modelscope/hub/utils/caching.py

@@ -1,9 +1,7 @@
 import hashlib
-import logging
 import os
 import pickle
 import tempfile
-import time
 from shutil import move, rmtree
 
 from modelscope.utils.logger import get_logger

modelscope/models/__init__.py

@@ -1,15 +1,30 @@
 # Copyright (c) Alibaba, Inc. and its affiliates.
-
-from .audio.ans.frcrn import FRCRNModel
-from .audio.kws import GenericKeyWordSpotting
-from .audio.tts.am import SambertNetHifi16k
-from .audio.tts.vocoder import Hifigan16k
 from .base import Model
 from .builder import MODELS, build_model
-from .multi_modal import OfaForImageCaptioning
-from .nlp import (BertForMaskedLM, BertForSequenceClassification, SbertForNLI,
-                  SbertForSentenceSimilarity, SbertForSentimentClassification,
-                  SbertForTokenClassification, SbertForZeroShotClassification,
-                  SpaceForDialogIntent, SpaceForDialogModeling,
-                  SpaceForDialogStateTracking, StructBertForMaskedLM,
-                  VecoForMaskedLM)
+
+try:
+    from .audio.tts.am import SambertNetHifi16k
+    from .audio.tts.vocoder import Hifigan16k
+
+except ModuleNotFoundError as e:
+    if str(e) == "No module named 'tensorflow'":
+        pass
+    else:
+        raise ModuleNotFoundError(e)
+
+try:
+    from .audio.kws import GenericKeyWordSpotting
+    from .multi_modal import OfaForImageCaptioning
+    from .nlp import (BertForMaskedLM, BertForSequenceClassification,
+                      SbertForNLI, SbertForSentenceSimilarity,
+                      SbertForSentimentClassification,
+                      SbertForTokenClassification,
+                      SbertForZeroShotClassification, SpaceForDialogIntent,
+                      SpaceForDialogModeling, SpaceForDialogStateTracking,
+                      StructBertForMaskedLM, VecoForMaskedLM)
+    from .audio.ans.frcrn import FRCRNModel
+except ModuleNotFoundError as e:
+    if str(e) == "No module named 'pytorch'":
+        pass
+    else:
+        raise ModuleNotFoundError(e)

modelscope/models/multi_modal/clip/clip_model.py

@@ -108,7 +108,11 @@ class CLIPForMultiModalEmbedding(Model):
         return text_ids_tensor, text_mask_tensor
 
     def forward(self, input: Dict[str, Any]) -> Dict[str, Any]:
-        output = {'img_embedding': None, 'text_embedding': None}
+        from modelscope.pipelines.outputs import OutputKeys
+        output = {
+            OutputKeys.IMG_EMBEDDING: None,
+            OutputKeys.TEXT_EMBEDDING: None
+        }
         if 'img' in input and input['img'] is not None:
             input_img = input['img']
             if isinstance(input_img, Image.Image):
@@ -130,7 +134,8 @@ class CLIPForMultiModalEmbedding(Model):
 
             img_embedding = self.clip_model(
                 input_data=img_tensor, input_type='img')
-            output['img_embedding'] = img_embedding.data.cpu().numpy()
+            from modelscope.pipelines.outputs import OutputKeys
+            output[OutputKeys.IMG_EMBEDDING] = img_embedding.data.cpu().numpy()
 
         if 'text' in input and input['text'] is not None:
             text_str = input['text']

modelscope/models/multi_modal/image_captioning_model.py

@@ -76,9 +76,10 @@ class OfaForImageCaptioning(Model):
             input = fairseq.utils.move_to_cuda(input, device=self._device)
         results, _ = self.eval_caption(self.task, self.generator, self.models,
                                        input)
+        from ...pipelines.outputs import OutputKeys
         return {
             'image_id': results[0]['image_id'],
-            'caption': results[0]['caption']
+            OutputKeys.CAPTION: results[0][OutputKeys.CAPTION]
         }
 
     def postprocess(self, inputs: Dict[str, Any]) -> Dict[str, Any]:

modelscope/msdatasets/config.py

@@ -2,6 +2,8 @@ import os
 from pathlib import Path
 
 # Cache location
+from modelscope.hub.constants import DEFAULT_MODELSCOPE_DATA_ENDPOINT
+
 DEFAULT_CACHE_HOME = '~/.cache'
 CACHE_HOME = os.getenv('CACHE_HOME', DEFAULT_CACHE_HOME)
 DEFAULT_MS_CACHE_HOME = os.path.join(CACHE_HOME, 'modelscope/hub')
@@ -18,5 +20,5 @@ DEFAULT_DOWNLOADED_DATASETS_PATH = os.path.join(MS_DATASETS_CACHE,
 DOWNLOADED_DATASETS_PATH = Path(
     os.getenv('DOWNLOADED_DATASETS_PATH', DEFAULT_DOWNLOADED_DATASETS_PATH))
 
-MS_HUB_ENDPOINT = os.environ.get('MS_HUB_ENDPOINT',
-                                 'http://47.94.223.21:31752')
+HUB_DATASET_ENDPOINT = os.environ.get('HUB_DATASET_ENDPOINT',
+                                      DEFAULT_MODELSCOPE_DATA_ENDPOINT)

modelscope/msdatasets/ms_dataset.py

@@ -11,7 +11,6 @@ from datasets.utils.file_utils import (is_relative_path,
                                        relative_to_absolute_path)
 
 from modelscope.msdatasets.config import MS_DATASETS_CACHE
-from modelscope.msdatasets.utils.ms_api import MsApi
 from modelscope.utils.constant import DownloadMode, Hubs
 from modelscope.utils.logger import get_logger
 
@@ -146,8 +145,9 @@ class MsDataset:
                 use_hf = True
             elif is_relative_path(dataset_name) and dataset_name.count(
                     '/') == 0:
-                ms_api = MsApi()
-                dataset_scripts = ms_api.fetch_dataset_scripts(
+                from modelscope.hub.api import HubApi
+                api = HubApi()
+                dataset_scripts = api.fetch_dataset_scripts(
                     dataset_name, namespace, download_mode, version)
                 if 'py' in dataset_scripts:  # dataset copied from hf datasets
                     dataset_name = dataset_scripts['py'][0]

modelscope/msdatasets/utils/ms_api.py

@@ -1,84 +0,0 @@
-import os
-import shutil
-from collections import defaultdict
-from typing import Optional
-
-import requests
-
-from modelscope.hub.errors import NotExistError, datahub_raise_on_error
-from modelscope.msdatasets.config import (DOWNLOADED_DATASETS_PATH,
-                                          MS_HUB_ENDPOINT)
-from modelscope.utils.constant import DownloadMode
-from modelscope.utils.logger import get_logger
-
-logger = get_logger()
-
-
-class MsApi:
-
-    def __init__(self, endpoint=MS_HUB_ENDPOINT):
-        self.endpoint = endpoint
-
-    def list_datasets(self):
-        path = f'{self.endpoint}/api/v1/datasets'
-        headers = None
-        params = {}
-        r = requests.get(path, params=params, headers=headers)
-        r.raise_for_status()
-        dataset_list = r.json()['Data']
-        return [x['Name'] for x in dataset_list]
-
-    def fetch_dataset_scripts(self,
-                              dataset_name: str,
-                              namespace: str,
-                              download_mode: Optional[DownloadMode],
-                              version: Optional[str] = 'master'):
-        if namespace is None:
-            raise ValueError(
-                f'Dataset from Hubs.modelscope should have a valid "namespace", but get {namespace}'
-            )
-        version = version or 'master'
-        cache_dir = os.path.join(DOWNLOADED_DATASETS_PATH, dataset_name,
-                                 namespace, version)
-        download_mode = DownloadMode(download_mode
-                                     or DownloadMode.REUSE_DATASET_IF_EXISTS)
-        if download_mode == DownloadMode.FORCE_REDOWNLOAD and os.path.exists(
-                cache_dir):
-            shutil.rmtree(cache_dir)
-        os.makedirs(cache_dir, exist_ok=True)
-        datahub_url = f'{self.endpoint}/api/v1/datasets/{namespace}/{dataset_name}'
-        r = requests.get(datahub_url)
-        resp = r.json()
-        datahub_raise_on_error(datahub_url, resp)
-        dataset_id = resp['Data']['Id']
-        datahub_url = f'{self.endpoint}/api/v1/datasets/{dataset_id}/repo/tree?Revision={version}'
-        r = requests.get(datahub_url)
-        resp = r.json()
-        datahub_raise_on_error(datahub_url, resp)
-        file_list = resp['Data']
-        if file_list is None:
-            raise NotExistError(
-                f'The modelscope dataset [dataset_name = {dataset_name}, namespace = {namespace}, '
-                f'version = {version}] dose not exist')
-
-        file_list = file_list['Files']
-        local_paths = defaultdict(list)
-        for file_info in file_list:
-            file_path = file_info['Path']
-            if file_path.endswith('.py'):
-                datahub_url = f'{self.endpoint}/api/v1/datasets/{dataset_id}/repo/files?' \
-                              f'Revision={version}&Path={file_path}'
-                r = requests.get(datahub_url)
-                r.raise_for_status()
-                content = r.json()['Data']['Content']
-                local_path = os.path.join(cache_dir, file_path)
-                if os.path.exists(local_path):
-                    logger.warning(
-                        f"Reusing dataset {dataset_name}'s python file ({local_path})"
-                    )
-                    local_paths['py'].append(local_path)
-                    continue
-                with open(local_path, 'w') as f:
-                    f.writelines(content)
-                local_paths['py'].append(local_path)
-        return local_paths

modelscope/pipelines/audio/__init__.py

@@ -1,3 +1,16 @@
-from .kws_kwsbp_pipeline import *  # noqa F403
-from .linear_aec_pipeline import LinearAECPipeline
-from .text_to_speech_pipeline import *  # noqa F403
+try:
+    from .kws_kwsbp_pipeline import *  # noqa F403
+    from .linear_aec_pipeline import LinearAECPipeline
+except ModuleNotFoundError as e:
+    if str(e) == "No module named 'torch'":
+        pass
+    else:
+        raise ModuleNotFoundError(e)
+
+try:
+    from .text_to_speech_pipeline import *  # noqa F403
+except ModuleNotFoundError as e:
+    if str(e) == "No module named 'tensorflow'":
+        pass
+    else:
+        raise ModuleNotFoundError(e)

modelscope/pipelines/audio/ans_pipeline.py

@@ -10,6 +10,7 @@ from modelscope.metainfo import Pipelines
 from modelscope.utils.constant import Tasks
 from ..base import Input, Pipeline
 from ..builder import PIPELINES
+from ..outputs import OutputKeys
 
 
 def audio_norm(x):
@@ -108,10 +109,10 @@ class ANSPipeline(Pipeline):
                     current_idx += stride
             else:
                 outputs = self.model(ndarray)['wav_l2'][0].cpu().numpy()
-        return {'output_pcm': outputs[:nsamples]}
+        return {OutputKeys.OUTPUT_PCM: outputs[:nsamples]}
 
     def postprocess(self, inputs: Dict[str, Any], **kwargs) -> Dict[str, Any]:
         if 'output_path' in kwargs.keys():
-            sf.write(kwargs['output_path'], inputs['output_pcm'],
+            sf.write(kwargs['output_path'], inputs[OutputKeys.OUTPUT_PCM],
                      self.SAMPLE_RATE)
         return inputs

modelscope/pipelines/audio/kws_kwsbp_pipeline.py

@@ -5,6 +5,8 @@ import stat
 import subprocess
 from typing import Any, Dict, List
 
+import json
+
 from modelscope.metainfo import Pipelines
 from modelscope.models import Model
 from modelscope.pipelines.base import Pipeline
@@ -39,6 +41,12 @@ class KeyWordSpottingKwsbpPipeline(Pipeline):
 
         self._preprocessor = preprocessor
         self._model = model
+        self._keywords = None
+
+        if 'keywords' in kwargs.keys():
+            self._keywords = kwargs['keywords']
+            print('self._keywords len: ', len(self._keywords))
+            print('self._keywords: ', self._keywords)
 
     def __call__(self, kws_type: str, wav_path: List[str]) -> Dict[str, Any]:
         assert kws_type in ['wav', 'pos_testsets', 'neg_testsets',
@@ -197,6 +205,16 @@ class KeyWordSpottingKwsbpPipeline(Pipeline):
         return rst_dict
 
     def _run_with_kwsbp(self, inputs: Dict[str, Any]) -> Dict[str, Any]:
+        opts: str = ''
+
+        # setting customized keywords
+        keywords_json = self._set_customized_keywords()
+        if len(keywords_json) > 0:
+            keywords_json_file = os.path.join(inputs['workspace'],
+                                              'keyword_custom.json')
+            with open(keywords_json_file, 'w') as f:
+                json.dump(keywords_json, f)
+            opts = '--keyword-custom ' + keywords_json_file
 
         if inputs['kws_set'] == 'roc':
             inputs['keyword_grammar_path'] = os.path.join(
@@ -211,7 +229,7 @@ class KeyWordSpottingKwsbpPipeline(Pipeline):
                 ' --sample-rate=' + inputs['sample_rate'] + \
                 ' --keyword-grammar=' + inputs['keyword_grammar_path'] + \
                 ' --wave-scp=' + os.path.join(inputs['pos_data_path'], 'wave.list') + \
-                ' --num-thread=1 > ' + dump_log_path + ' 2>&1'
+                ' --num-thread=1 ' + opts + ' > ' + dump_log_path + ' 2>&1'
             os.system(kws_cmd)
 
         if inputs['kws_set'] in ['pos_testsets', 'roc']:
@@ -236,7 +254,7 @@ class KeyWordSpottingKwsbpPipeline(Pipeline):
                     ' --sample-rate=' + inputs['sample_rate'] + \
                     ' --keyword-grammar=' + inputs['keyword_grammar_path'] + \
                     ' --wave-scp=' + wav_list_path + \
-                    ' --num-thread=1 > ' + dump_log_path + ' 2>&1'
+                    ' --num-thread=1 ' + opts + ' > ' + dump_log_path + ' 2>&1'
                 p = subprocess.Popen(kws_cmd, shell=True)
                 process.append(p)
                 j += 1
@@ -268,7 +286,7 @@ class KeyWordSpottingKwsbpPipeline(Pipeline):
                     ' --sample-rate=' + inputs['sample_rate'] + \
                     ' --keyword-grammar=' + inputs['keyword_grammar_path'] + \
                     ' --wave-scp=' + wav_list_path + \
-                    ' --num-thread=1 > ' + dump_log_path + ' 2>&1'
+                    ' --num-thread=1 ' + opts + ' > ' + dump_log_path + ' 2>&1'
                 p = subprocess.Popen(kws_cmd, shell=True)
                 process.append(p)
                 j += 1
@@ -447,3 +465,29 @@ class KeyWordSpottingKwsbpPipeline(Pipeline):
             threshold_cur += step
 
         return output
+
+    def _set_customized_keywords(self) -> Dict[str, Any]:
+        if self._keywords is not None:
+            word_list_inputs = self._keywords
+            word_list = []
+            for i in range(len(word_list_inputs)):
+                key = word_list_inputs[i]
+                new_item = {}
+                if key.__contains__('keyword'):
+                    name = key['keyword']
+                    new_name: str = ''
+                    for n in range(0, len(name), 1):
+                        new_name += name[n]
+                        new_name += ' '
+                    new_name = new_name.strip()
+                    new_item['name'] = new_name
+
+                    if key.__contains__('threshold'):
+                        threshold1: float = key['threshold']
+                        new_item['threshold1'] = threshold1
+
+                word_list.append(new_item)
+            out = {'word_list': word_list}
+            return out
+        else:
+            return ''

modelscope/pipelines/audio/linear_aec_pipeline.py

@@ -12,6 +12,7 @@ from modelscope.preprocessors.audio import LinearAECAndFbank
 from modelscope.utils.constant import ModelFile, Tasks
 from ..base import Pipeline
 from ..builder import PIPELINES
+from ..outputs import OutputKeys
 
 FEATURE_MVN = 'feature.DEY.mvn.txt'
 
@@ -120,7 +121,7 @@ class LinearAECPipeline(Pipeline):
                 }
         """
         output_data = self._process(inputs['feature'], inputs['base'])
-        return {'output_pcm': output_data}
+        return {OutputKeys.OUTPUT_PCM: output_data}
 
     def postprocess(self, inputs: Dict[str, Any], **kwargs) -> Dict[str, Any]:
         r"""The post process. Will save audio to file, if the output_path is given.
@@ -140,8 +141,8 @@ class LinearAECPipeline(Pipeline):
         """
         if 'output_path' in kwargs.keys():
             wav.write(kwargs['output_path'], self.preprocessor.SAMPLE_RATE,
-                      inputs['output_pcm'].astype(np.int16))
-        inputs['output_pcm'] = inputs['output_pcm'] / 32768.0
+                      inputs[OutputKeys.OUTPUT_PCM].astype(np.int16))
+        inputs[OutputKeys.OUTPUT_PCM] = inputs[OutputKeys.OUTPUT_PCM] / 32768.0
         return inputs
 
     def _process(self, fbanks, mixture):

modelscope/pipelines/cv/__init__.py

@@ -1,5 +1,18 @@
-from .action_recognition_pipeline import ActionRecognitionPipeline
-from .animal_recog_pipeline import AnimalRecogPipeline
-from .image_cartoon_pipeline import ImageCartoonPipeline
-from .image_matting_pipeline import ImageMattingPipeline
-from .ocr_detection_pipeline import OCRDetectionPipeline
+try:
+    from .action_recognition_pipeline import ActionRecognitionPipeline
+    from .animal_recog_pipeline import AnimalRecogPipeline
+except ModuleNotFoundError as e:
+    if str(e) == "No module named 'torch'":
+        pass
+    else:
+        raise ModuleNotFoundError(e)
+
+try:
+    from .image_cartoon_pipeline import ImageCartoonPipeline
+    from .image_matting_pipeline import ImageMattingPipeline
+    from .ocr_detection_pipeline import OCRDetectionPipeline
+except ModuleNotFoundError as e:
+    if str(e) == "No module named 'tensorflow'":
+        pass
+    else:
+        raise ModuleNotFoundError(e)

modelscope/pipelines/cv/action_recognition_pipeline.py

@@ -16,6 +16,7 @@ from modelscope.utils.constant import ModelFile, Tasks
 from modelscope.utils.logger import get_logger
 from ..base import Pipeline
 from ..builder import PIPELINES
+from ..outputs import OutputKeys
 
 logger = get_logger()
 
@@ -49,7 +50,7 @@ class ActionRecognitionPipeline(Pipeline):
     def forward(self, input: Dict[str, Any]) -> Dict[str, Any]:
         pred = self.perform_inference(input['video_data'])
         output_label = self.label_mapping[str(pred)]
-        return {'output_label': output_label}
+        return {OutputKeys.LABELS: output_label}
 
     @torch.no_grad()
     def perform_inference(self, data, max_bsz=4):

modelscope/pipelines/cv/animal_recog_pipeline.py

@@ -18,6 +18,7 @@ from modelscope.utils.constant import ModelFile, Tasks
 from modelscope.utils.logger import get_logger
 from ..base import Pipeline
 from ..builder import PIPELINES
+from ..outputs import OutputKeys
 
 logger = get_logger()
 
@@ -121,7 +122,9 @@ class AnimalRecogPipeline(Pipeline):
             label_mapping = f.readlines()
         score = torch.max(inputs['outputs'])
         inputs = {
-            'scores': score.item(),
-            'labels': label_mapping[inputs['outputs'].argmax()].split('\t')[1]
+            OutputKeys.SCORES:
+            score.item(),
+            OutputKeys.LABELS:
+            label_mapping[inputs['outputs'].argmax()].split('\t')[1]
         }
         return inputs

modelscope/pipelines/cv/image_cartoon_pipeline.py

@@ -17,6 +17,7 @@ from modelscope.utils.constant import Tasks
 from modelscope.utils.logger import get_logger
 from ..base import Pipeline
 from ..builder import PIPELINES
+from ..outputs import OutputKeys
 
 if tf.__version__ >= '2.0':
     tf = tf.compat.v1
@@ -94,7 +95,7 @@ class ImageCartoonPipeline(Pipeline):
         landmarks = self.detect_face(img)
         if landmarks is None:
             print('No face detected!')
-            return {'output_png': None}
+            return {OutputKeys.OUTPUT_IMG: None}
 
         # background process
         pad_bg, pad_h, pad_w = padTo16x(img_brg)
@@ -143,7 +144,7 @@ class ImageCartoonPipeline(Pipeline):
 
         res = cv2.resize(res, (ori_w, ori_h), interpolation=cv2.INTER_AREA)
 
-        return {'output_png': res}
+        return {OutputKeys.OUTPUT_IMG: res}
 
     def postprocess(self, inputs: Dict[str, Any]) -> Dict[str, Any]:
         return inputs

modelscope/pipelines/cv/image_matting_pipeline.py

@@ -12,6 +12,7 @@ from modelscope.utils.constant import ModelFile, Tasks
 from modelscope.utils.logger import get_logger
 from ..base import Pipeline
 from ..builder import PIPELINES
+from ..outputs import OutputKeys
 
 logger = get_logger()
 
@@ -60,9 +61,9 @@ class ImageMattingPipeline(Pipeline):
     def forward(self, input: Dict[str, Any]) -> Dict[str, Any]:
         with self._session.as_default():
             feed_dict = {self.input_name: input['img']}
-            output_png = self._session.run(self.output, feed_dict=feed_dict)
-            output_png = cv2.cvtColor(output_png, cv2.COLOR_RGBA2BGRA)
-            return {'output_png': output_png}
+            output_img = self._session.run(self.output, feed_dict=feed_dict)
+            output_img = cv2.cvtColor(output_img, cv2.COLOR_RGBA2BGRA)
+            return {OutputKeys.OUTPUT_IMG: output_img}
 
     def postprocess(self, inputs: Dict[str, Any]) -> Dict[str, Any]:
         return inputs

modelscope/pipelines/cv/ocr_detection_pipeline.py

@@ -16,6 +16,7 @@ from modelscope.utils.constant import ModelFile, Tasks
 from modelscope.utils.logger import get_logger
 from ..base import Pipeline
 from ..builder import PIPELINES
+from ..outputs import OutputKeys
 from .ocr_utils import model_resnet_mutex_v4_linewithchar, ops, utils
 
 if tf.__version__ >= '2.0':
@@ -174,5 +175,5 @@ class OCRDetectionPipeline(Pipeline):
         dt_nms = utils.nms_python(dt_n9)
         dt_polygons = np.array([o[:8] for o in dt_nms])
 
-        result = {'det_polygons': dt_polygons}
+        result = {OutputKeys.POLYGONS: dt_polygons}
         return result

modelscope/pipelines/multi_modal/__init__.py

@@ -1,3 +1,9 @@
-from .image_captioning_pipeline import ImageCaptionPipeline
-from .multi_modal_embedding_pipeline import MultiModalEmbeddingPipeline
-from .visual_question_answering_pipeline import VisualQuestionAnsweringPipeline
+try:
+    from .image_captioning_pipeline import ImageCaptionPipeline
+    from .multi_modal_embedding_pipeline import MultiModalEmbeddingPipeline
+    from .visual_question_answering_pipeline import VisualQuestionAnsweringPipeline
+except ModuleNotFoundError as e:
+    if str(e) == "No module named 'torch'":
+        pass
+    else:
+        raise ModuleNotFoundError(e)

modelscope/pipelines/nlp/__init__.py

@@ -1,11 +1,17 @@
-from .dialog_intent_prediction_pipeline import *  # noqa F403
-from .dialog_modeling_pipeline import *  # noqa F403
-from .dialog_state_tracking_pipeline import *  # noqa F403
-from .fill_mask_pipeline import *  # noqa F403
-from .nli_pipeline import *  # noqa F403
-from .sentence_similarity_pipeline import *  # noqa F403
-from .sentiment_classification_pipeline import *  # noqa F403
-from .sequence_classification_pipeline import *  # noqa F403
-from .text_generation_pipeline import *  # noqa F403
-from .word_segmentation_pipeline import *  # noqa F403
-from .zero_shot_classification_pipeline import *  # noqa F403
+try:
+    from .dialog_intent_prediction_pipeline import *  # noqa F403
+    from .dialog_modeling_pipeline import *  # noqa F403
+    from .dialog_state_tracking_pipeline import *  # noqa F403
+    from .fill_mask_pipeline import *  # noqa F403
+    from .nli_pipeline import *  # noqa F403
+    from .sentence_similarity_pipeline import *  # noqa F403
+    from .sentiment_classification_pipeline import *  # noqa F403
+    from .sequence_classification_pipeline import *  # noqa F403
+    from .text_generation_pipeline import *  # noqa F403
+    from .word_segmentation_pipeline import *  # noqa F403
+    from .zero_shot_classification_pipeline import *  # noqa F403
+except ModuleNotFoundError as e:
+    if str(e) == "No module named 'torch'":
+        pass
+    else:
+        raise ModuleNotFoundError(e)

modelscope/pipelines/nlp/dialog_intent_prediction_pipeline.py

@@ -8,6 +8,7 @@ from ...preprocessors import DialogIntentPredictionPreprocessor
 from ...utils.constant import Tasks
 from ..base import Pipeline
 from ..builder import PIPELINES
+from ..outputs import OutputKeys
 
 __all__ = ['DialogIntentPredictionPipeline']
 
@@ -44,9 +45,9 @@ class DialogIntentPredictionPipeline(Pipeline):
         pos = np.where(pred == np.max(pred))
 
         result = {
-            'prediction': pred,
-            'label_pos': pos[0],
-            'label': self.categories[pos[0][0]]
+            OutputKeys.PREDICTION: pred,
+            OutputKeys.LABEL_POS: pos[0],
+            OutputKeys.LABEL: self.categories[pos[0][0]]
         }
 
         return result

modelscope/pipelines/nlp/dialog_modeling_pipeline.py

@@ -8,6 +8,7 @@ from ...preprocessors import DialogModelingPreprocessor
 from ...utils.constant import Tasks
 from ..base import Pipeline, Tensor
 from ..builder import PIPELINES
+from ..outputs import OutputKeys
 
 __all__ = ['DialogModelingPipeline']
 
@@ -42,7 +43,6 @@ class DialogModelingPipeline(Pipeline):
             inputs['resp'])
         assert len(sys_rsp) > 2
         sys_rsp = sys_rsp[1:len(sys_rsp) - 1]
-
-        inputs['response'] = sys_rsp
+        inputs[OutputKeys.RESPONSE] = sys_rsp
 
         return inputs

modelscope/pipelines/nlp/fill_mask_pipeline.py

@@ -11,6 +11,7 @@ from ...utils.config import Config
 from ...utils.constant import ModelFile, Tasks
 from ..base import Pipeline, Tensor
 from ..builder import PIPELINES
+from ..outputs import OutputKeys
 
 __all__ = ['FillMaskPipeline']
 _type_map = {'veco': 'roberta', 'sbert': 'bert'}
@@ -108,4 +109,4 @@ class FillMaskPipeline(Pipeline):
             pred_string = rep_tokens(pred_string, self.rep_map[process_type])
             pred_strings.append(pred_string)
 
-        return {'text': pred_strings}
+        return {OutputKeys.TEXT: pred_strings}

modelscope/pipelines/nlp/nli_pipeline.py

@@ -11,6 +11,7 @@ from ...preprocessors import NLIPreprocessor
 from ...utils.constant import Tasks
 from ..base import Pipeline
 from ..builder import PIPELINES
+from ..outputs import OutputKeys
 
 __all__ = ['NLIPipeline']
 
@@ -69,4 +70,4 @@ class NLIPipeline(Pipeline):
 
         cls_names = [self.model.id2label[cid] for cid in cls_ids]
 
-        return {'scores': probs, 'labels': cls_names}
+        return {OutputKeys.SCORES: probs, OutputKeys.LABELS: cls_names}

modelscope/pipelines/nlp/sentence_similarity_pipeline.py

@@ -10,6 +10,7 @@ from ...preprocessors import SequenceClassificationPreprocessor
 from ...utils.constant import Tasks
 from ..base import Input, Pipeline
 from ..builder import PIPELINES
+from ..outputs import OutputKeys
 
 __all__ = ['SentenceSimilarityPipeline']
 
@@ -69,4 +70,4 @@ class SentenceSimilarityPipeline(Pipeline):
         probs = probs[cls_ids].tolist()
         cls_names = [self.model.id2label[cid] for cid in cls_ids]
         b = 0
-        return {'scores': probs[b], 'labels': cls_names[b]}
+        return {OutputKeys.SCORES: probs[b], OutputKeys.LABELS: cls_names[b]}

modelscope/pipelines/nlp/sentiment_classification_pipeline.py

@@ -13,6 +13,7 @@ from ...preprocessors import SentimentClassificationPreprocessor
 from ...utils.constant import Tasks
 from ..base import Input, Pipeline
 from ..builder import PIPELINES
+from ..outputs import OutputKeys
 
 __all__ = ['SentimentClassificationPipeline']
 
@@ -73,5 +74,4 @@ class SentimentClassificationPipeline(Pipeline):
         probs = probs[cls_ids].tolist()
 
         cls_names = [self.model.id2label[cid] for cid in cls_ids]
-
-        return {'scores': probs, 'labels': cls_names}
+        return {OutputKeys.SCORES: probs, OutputKeys.LABELS: cls_names}

modelscope/pipelines/nlp/sequence_classification_pipeline.py

@@ -9,6 +9,7 @@ from modelscope.utils.constant import Tasks
 from ...models import Model
 from ..base import Input, Pipeline
 from ..builder import PIPELINES
+from ..outputs import OutputKeys
 
 __all__ = ['SequenceClassificationPipeline']
 
@@ -64,4 +65,4 @@ class SequenceClassificationPipeline(Pipeline):
 
         cls_names = [self.model.id2label[cid] for cid in cls_ids]
 
-        return {'scores': probs, 'labels': cls_names}
+        return {OutputKeys.SCORES: probs, OutputKeys.LABELS: cls_names}

modelscope/pipelines/nlp/text_generation_pipeline.py

@@ -9,6 +9,7 @@ from ...preprocessors import TextGenerationPreprocessor
 from ...utils.constant import Tasks
 from ..base import Pipeline, Tensor
 from ..builder import PIPELINES
+from ..outputs import OutputKeys
 
 __all__ = ['TextGenerationPipeline']
 
@@ -70,4 +71,4 @@ class TextGenerationPipeline(Pipeline):
         for _old, _new in replace_tokens_roberta:
             pred_string = pred_string.replace(_old, _new)
         pred_string.strip()
-        return {'text': pred_string}
+        return {OutputKeys.TEXT: pred_string}

modelscope/pipelines/nlp/word_segmentation_pipeline.py

@@ -9,6 +9,7 @@ from ...preprocessors import TokenClassifcationPreprocessor
 from ...utils.constant import Tasks
 from ..base import Pipeline, Tensor
 from ..builder import PIPELINES
+from ..outputs import OutputKeys
 
 __all__ = ['WordSegmentationPipeline']
 
@@ -73,7 +74,4 @@ class WordSegmentationPipeline(Pipeline):
         if chunk:
             chunks.append(chunk)
         seg_result = ' '.join(chunks)
-        rst = {
-            'output': seg_result,
-        }
-        return rst
+        return {OutputKeys.OUTPUT: seg_result}

modelscope/pipelines/nlp/zero_shot_classification_pipeline.py

@@ -14,6 +14,7 @@ from ...preprocessors import ZeroShotClassificationPreprocessor
 from ...utils.constant import Tasks
 from ..base import Input, Pipeline
 from ..builder import PIPELINES
+from ..outputs import OutputKeys
 
 __all__ = ['ZeroShotClassificationPipeline']
 
@@ -82,7 +83,7 @@ class ZeroShotClassificationPipeline(Pipeline):
             scores = softmax(logits, axis=-1)
         reversed_index = list(reversed(scores.argsort()))
         result = {
-            'labels': [candidate_labels[i] for i in reversed_index],
-            'scores': [scores[i].item() for i in reversed_index]
+            OutputKeys.LABELS: [candidate_labels[i] for i in reversed_index],
+            OutputKeys.SCORES: [scores[i].item() for i in reversed_index],
         }
         return result

modelscope/pipelines/outputs.py

@@ -2,54 +2,76 @@
 
 from modelscope.utils.constant import Tasks
 
+
+class OutputKeys(object):
+    SCORES = 'scores'
+    LABEL = 'label'
+    LABELS = 'labels'
+    LABEL_POS = 'label_pos'
+    POSES = 'poses'
+    CAPTION = 'caption'
+    BOXES = 'boxes'
+    TEXT = 'text'
+    POLYGONS = 'polygons'
+    OUTPUT = 'output'
+    OUTPUT_IMG = 'output_img'
+    OUTPUT_PCM = 'output_pcm'
+    IMG_EMBEDDING = 'img_embedding'
+    TEXT_EMBEDDING = 'text_embedding'
+    RESPONSE = 'response'
+    PREDICTION = 'prediction'
+
+
 TASK_OUTPUTS = {
 
     # ============ vision tasks ===================
 
     # image classification result for single sample
     #   {
-    #       "labels": ["dog", "horse", "cow", "cat"],
     #       "scores": [0.9, 0.1, 0.05, 0.05]
+    #       "labels": ["dog", "horse", "cow", "cat"],
     #   }
-    Tasks.image_classification: ['scores', 'labels'],
-    Tasks.image_tagging: ['scores', 'labels'],
+    Tasks.image_classification: [OutputKeys.SCORES, OutputKeys.LABELS],
+    Tasks.image_tagging: [OutputKeys.SCORES, OutputKeys.LABELS],
 
     # object detection result for single sample
     #   {
+    #       "scores": [0.9, 0.1, 0.05, 0.05]
+    #       "labels": ["dog", "horse", "cow", "cat"],
     #       "boxes": [
     #           [x1, y1, x2, y2],
     #           [x1, y1, x2, y2],
     #           [x1, y1, x2, y2],
     #       ],
-    #       "labels": ["dog", "horse", "cow", "cat"],
-    #       "scores": [0.9, 0.1, 0.05, 0.05]
     #   }
-    Tasks.object_detection: ['scores', 'labels', 'boxes'],
+    Tasks.object_detection:
+    [OutputKeys.SCORES, OutputKeys.LABELS, OutputKeys.BOXES],
 
     # instance segmentation result for single sample
     #   {
-    #       "masks": [
-    #           np.array in bgr channel order
-    #       ],
+    #       "scores": [0.9, 0.1, 0.05, 0.05],
     #       "labels": ["dog", "horse", "cow", "cat"],
-    #       "scores": [0.9, 0.1, 0.05, 0.05]
+    #       "boxes": [
+    #           np.array in bgr channel order
+    #       ]
     #   }
-    Tasks.image_segmentation: ['scores', 'labels', 'boxes'],
+    Tasks.image_segmentation:
+    [OutputKeys.SCORES, OutputKeys.LABELS, OutputKeys.BOXES],
 
     # image generation/editing/matting result for single sample
     # {
-    #   "output_png": np.array with shape(h, w, 4)
+    #   "output_img": np.array with shape(h, w, 4)
     #                 for matting or (h, w, 3) for general purpose
     # }
-    Tasks.image_editing: ['output_png'],
-    Tasks.image_matting: ['output_png'],
-    Tasks.image_generation: ['output_png'],
+    Tasks.image_editing: [OutputKeys.OUTPUT_IMG],
+    Tasks.image_matting: [OutputKeys.OUTPUT_IMG],
+    Tasks.image_generation: [OutputKeys.OUTPUT_IMG],
 
     # action recognition result for single video
     # {
     #   "output_label": "abseiling"
     # }
-    Tasks.action_recognition: ['output_label'],
+    Tasks.action_recognition: [OutputKeys.LABELS],
 
     # pose estimation result for single sample
     # {
@@ -58,48 +80,55 @@ TASK_OUTPUTS = {
     #   "boxes": np.array with shape [num_pose, 4], each box is
     #       [x1, y1, x2, y2]
     # }
-    Tasks.pose_estimation: ['poses', 'boxes'],
+    Tasks.pose_estimation: [OutputKeys.POSES, OutputKeys.BOXES],
 
     # ocr detection result for single sample
     # {
-    #   "det_polygons": np.array with shape [num_text, 8], each box is
+    #   "polygons": np.array with shape [num_text, 8], each polygon is
     #       [x1, y1, x2, y2, x3, y3, x4, y4]
     # }
-    Tasks.ocr_detection: ['det_polygons'],
+    Tasks.ocr_detection: [OutputKeys.POLYGONS],
 
     # ============ nlp tasks ===================
 
     # text classification result for single sample
     #   {
-    #       "labels": ["happy", "sad", "calm", "angry"],
     #       "scores": [0.9, 0.1, 0.05, 0.05]
+    #       "labels": ["happy", "sad", "calm", "angry"],
     #   }
-    Tasks.text_classification: ['scores', 'labels'],
+    Tasks.text_classification: [OutputKeys.SCORES, OutputKeys.LABELS],
 
     # text generation result for single sample
     # {
-    #   "text": "this is text generated by a model."
+    #   "text": "this is the text generated by a model."
     # }
-    Tasks.text_generation: ['text'],
+    Tasks.text_generation: [OutputKeys.TEXT],
 
     # fill mask result for single sample
     # {
     #   "text": "this is the text which masks filled by model."
     # }
-    Tasks.fill_mask: ['text'],
+    Tasks.fill_mask: [OutputKeys.TEXT],
 
     # word segmentation result for single sample
     # {
     #   "output": "今天 天气 不错 ， 适合 出去 游玩"
     # }
-    Tasks.word_segmentation: ['output'],
+    Tasks.word_segmentation: [OutputKeys.OUTPUT],
 
     # sentence similarity result for single sample
     #   {
-    #       "labels": "1",
     #       "scores": 0.9
+    #       "labels": "1",
     #   }
-    Tasks.sentence_similarity: ['scores', 'labels'],
+    Tasks.sentence_similarity: [OutputKeys.SCORES, OutputKeys.LABELS],
+
+    # sentiment classification result for single sample
+    #   {
+    #       "labels": ["happy", "sad", "calm", "angry"],
+    #       "scores": [0.9, 0.1, 0.05, 0.05]
+    #   }
+    Tasks.sentiment_classification: [OutputKeys.SCORES, OutputKeys.LABELS],
 
     # sentiment classification result for single sample
     #   {
@@ -110,10 +139,43 @@ TASK_OUTPUTS = {
 
     # zero-shot classification result for single sample
     #   {
+    #       "scores": [0.9, 0.1, 0.05, 0.05]
+    #       "labels": ["happy", "sad", "calm", "angry"],
+    #   }
+    Tasks.zero_shot_classification: [OutputKeys.SCORES, OutputKeys.LABELS],
+
+    # nli result for single sample
+    #   {
     #       "labels": ["happy", "sad", "calm", "angry"],
     #       "scores": [0.9, 0.1, 0.05, 0.05]
     #   }
-    Tasks.zero_shot_classification: ['scores', 'labels'],
+    Tasks.nli: [OutputKeys.SCORES, OutputKeys.LABELS],
+
+    # {'pred': array([2.62349960e-03, 4.12110658e-03, 4.12748595e-05, 3.77560973e-05,
+    #        1.08599677e-04, 1.72710388e-05, 2.95618793e-05, 1.93638436e-04,
+    #        6.45841064e-05, 1.15997791e-04, 5.11605394e-05, 9.87020373e-01,
+    #        2.66957268e-05, 4.72324500e-05, 9.74208378e-05, 4.18022355e-05,
+    #        2.97343540e-05, 5.81317654e-05, 5.44203431e-05, 6.28319322e-05,
+    #        7.34537680e-05, 6.61411541e-05, 3.62534920e-05, 8.58885178e-05,
+    #        8.24327726e-05, 4.66077945e-05, 5.32869453e-05, 4.16190960e-05,
+    #        5.97518992e-05, 3.92273068e-05, 3.44069012e-05, 9.92335918e-05,
+    #        9.25978165e-05, 6.26462061e-05, 3.32317031e-05, 1.32061413e-03,
+    #        2.01607945e-05, 3.36636294e-05, 3.99156743e-05, 5.84108493e-05,
+    #        2.53432900e-05, 4.95731190e-04, 2.64443643e-05, 4.46992999e-05,
+    #        2.42672231e-05, 4.75615161e-05, 2.66230145e-05, 4.00083954e-05,
+    #        2.90536875e-04, 4.23891543e-05, 8.63691166e-05, 4.98188965e-05,
+    #        3.47019341e-05, 4.52718523e-05, 4.20905781e-05, 5.50173208e-05,
+    #        4.92360487e-05, 3.56021264e-05, 2.13957210e-05, 6.17428886e-05,
+    #        1.43893281e-04, 7.32152112e-05, 2.91354867e-04, 2.46623786e-05,
+    #        3.61441926e-05, 3.38475402e-05, 3.44323053e-05, 5.70138109e-05,
+    #        4.31488479e-05, 4.94503947e-05, 4.30105974e-05, 1.00963116e-04,
+    #        2.82062047e-05, 1.15582036e-04, 4.48261271e-05, 3.99339879e-05,
+    #        7.27692823e-05], dtype=float32), 'label_pos': array([11]), 'label': 'lost_or_stolen_card'}
+    Tasks.dialog_intent_prediction:
+    [OutputKeys.PREDICTION, OutputKeys.LABEL_POS, OutputKeys.LABEL],
+
+    # sys : ['you', 'are', 'welcome', '.', 'have', 'a', 'great', 'day', '!']
+    Tasks.dialog_modeling: [OutputKeys.RESPONSE],
 
     # nli result for single sample
     #   {
@@ -189,7 +251,7 @@ TASK_OUTPUTS = {
     # {
     #   "output_pcm": np.array with shape(samples,) and dtype float32
     # }
-    Tasks.speech_signal_process: ['output_pcm'],
+    Tasks.speech_signal_process: [OutputKeys.OUTPUT_PCM],
 
     # ============ multi-modal tasks ===================
 
@@ -197,14 +259,15 @@ TASK_OUTPUTS = {
     # {
     #   "caption": "this is an image caption text."
     # }
-    Tasks.image_captioning: ['caption'],
+    Tasks.image_captioning: [OutputKeys.CAPTION],
 
     # multi-modal embedding result for single sample
     # {
     #   "img_embedding": np.array with shape [1, D],
     #   "text_embedding": np.array with shape [1, D]
     # }
-    Tasks.multi_modal_embedding: ['img_embedding', 'text_embedding'],
+    Tasks.multi_modal_embedding:
+    [OutputKeys.IMG_EMBEDDING, OutputKeys.TEXT_EMBEDDING],
 
     # visual grounding result for single sample
     # {
@@ -215,11 +278,11 @@ TASK_OUTPUTS = {
     #       ],
     #       "scores": [0.9, 0.1, 0.05, 0.05]
     # }
-    Tasks.visual_grounding: ['boxes', 'scores'],
+    Tasks.visual_grounding: [OutputKeys.BOXES, OutputKeys.SCORES],
 
     # text_to_image result for a single sample
     # {
-    #    "image": np.ndarray with shape [height, width, 3]
+    #    "output_img": np.ndarray with shape [height, width, 3]
     # }
-    Tasks.text_to_image_synthesis: ['image']
+    Tasks.text_to_image_synthesis: [OutputKeys.OUTPUT_IMG]
 }

modelscope/preprocessors/__init__.py

@@ -1,14 +1,21 @@
 # Copyright (c) Alibaba, Inc. and its affiliates.
 
-from .audio import LinearAECAndFbank
 from .base import Preprocessor
 from .builder import PREPROCESSORS, build_preprocessor
 from .common import Compose
 from .image import LoadImage, load_image
 from .kws import WavToLists
-from .multi_modal import *  # noqa F403
-from .nlp import *  # noqa F403
-from .space.dialog_intent_prediction_preprocessor import *  # noqa F403
-from .space.dialog_modeling_preprocessor import *  # noqa F403
-from .space.dialog_state_tracking_preprocessor import *  # noqa F403
 from .text_to_speech import *  # noqa F403
+
+try:
+    from .audio import LinearAECAndFbank
+    from .multi_modal import *  # noqa F403
+    from .nlp import *  # noqa F403
+    from .space.dialog_intent_prediction_preprocessor import *  # noqa F403
+    from .space.dialog_modeling_preprocessor import *  # noqa F403
+    from .space.dialog_state_tracking_preprocessor import *  # noqa F403
+except ModuleNotFoundError as e:
+    if str(e) == "No module named 'tensorflow'":
+        pass
+    else:
+        raise ModuleNotFoundError(e)

modelscope/trainers/nlp/sequence_classification_trainer.py

@@ -14,8 +14,7 @@ PATH = None
 logger = get_logger(PATH)
 
 
-@TRAINERS.register_module(
-    Tasks.text_classification, module_name=r'bert-sentiment-analysis')
+@TRAINERS.register_module(module_name=r'bert-sentiment-analysis')
 class SequenceClassificationTrainer(BaseTrainer):
 
     def __init__(self, cfg_file: str, *args, **kwargs):

modelscope/utils/check_requirements.py

@@ -0,0 +1,79 @@
+# Copyright (c) Alibaba, Inc. and its affiliates.
+
+from modelscope.utils.constant import Fields, Requirements
+from modelscope.utils.import_utils import requires
+
+
+def get_msg(field):
+    msg = f'\n{field} requirements not installed, please execute ' \
+          f'`pip install requirements/{field}.txt` or ' \
+          f'`pip install modelscope[{field}]`'
+    return msg
+
+
+class NLPModuleNotFoundError(ModuleNotFoundError):
+
+    def __init__(self, e: ModuleNotFoundError) -> None:
+        e.msg += get_msg(Fields.nlp)
+        super().__init__(e)
+
+
+class CVModuleNotFoundError(ModuleNotFoundError):
+
+    def __init__(self, e: ModuleNotFoundError) -> None:
+        e.msg += get_msg(Fields.cv)
+        super().__init__(e)
+
+
+class AudioModuleNotFoundError(ModuleNotFoundError):
+
+    def __init__(self, e: ModuleNotFoundError) -> None:
+        e.msg += get_msg(Fields.audio)
+        super().__init__(e)
+
+
+class MultiModalModuleNotFoundError(ModuleNotFoundError):
+
+    def __init__(self, e: ModuleNotFoundError) -> None:
+        e.msg += get_msg(Fields.multi_modal)
+        super().__init__(e)
+
+
+def check_nlp():
+    try:
+        requires('nlp models', (
+            Requirements.torch,
+            Requirements.tokenizers,
+        ))
+    except ImportError as e:
+        raise NLPModuleNotFoundError(e)
+
+
+def check_cv():
+    try:
+        requires('cv models', (
+            Requirements.torch,
+            Requirements.tokenizers,
+        ))
+    except ImportError as e:
+        raise CVModuleNotFoundError(e)
+
+
+def check_audio():
+    try:
+        requires('audio models', (
+            Requirements.torch,
+            Requirements.tf,
+        ))
+    except ImportError as e:
+        raise AudioModuleNotFoundError(e)
+
+
+def check_multi_modal():
+    try:
+        requires('multi-modal models', (
+            Requirements.torch,
+            Requirements.tokenizers,
+        ))
+    except ImportError as e:
+        raise MultiModalModuleNotFoundError(e)

modelscope/utils/config.py

@@ -17,9 +17,10 @@ from typing import Dict
 import addict
 from yapf.yapflib.yapf_api import FormatCode
 
+from modelscope.utils.import_utils import (import_modules,
+                                           import_modules_from_file,
+                                           validate_py_syntax)
 from modelscope.utils.logger import get_logger
-from modelscope.utils.pymod import (import_modules, import_modules_from_file,
-                                    validate_py_syntax)
 
 if platform.system() == 'Windows':
     import regex as re  # type: ignore

modelscope/utils/constant.py

@@ -102,5 +102,18 @@ class ModelFile(object):
     TORCH_MODEL_BIN_FILE = 'pytorch_model.bin'
 
 
+class Requirements(object):
+    """Requirement names for each module
+    """
+    protobuf = 'protobuf'
+    sentencepiece = 'sentencepiece'
+    sklearn = 'sklearn'
+    scipy = 'scipy'
+    timm = 'timm'
+    tokenizers = 'tokenizers'
+    tf = 'tf'
+    torch = 'torch'
+
+
 TENSORFLOW = 'tensorflow'
 PYTORCH = 'pytorch'

modelscope/utils/import_utils.py

@@ -0,0 +1,324 @@
+# Copyright (c) Alibaba, Inc. and its affiliates.
+# Part of the implementation is borrowed from huggingface/transformers.
+import ast
+import functools
+import importlib.util
+import os
+import os.path as osp
+import sys
+import types
+from collections import OrderedDict
+from functools import wraps
+from importlib import import_module
+from itertools import chain
+from types import ModuleType
+from typing import Any
+
+import json
+from packaging import version
+
+from modelscope.utils.constant import Fields
+from modelscope.utils.logger import get_logger
+
+if sys.version_info < (3, 8):
+    import importlib_metadata
+else:
+    import importlib.metadata as importlib_metadata
+
+logger = get_logger()
+
+
+def import_modules_from_file(py_file: str):
+    """ Import module from a certrain file
+
+    Args:
+        py_file: path to a python file to be imported
+
+    Return:
+
+    """
+    dirname, basefile = os.path.split(py_file)
+    if dirname == '':
+        dirname == './'
+    module_name = osp.splitext(basefile)[0]
+    sys.path.insert(0, dirname)
+    validate_py_syntax(py_file)
+    mod = import_module(module_name)
+    sys.path.pop(0)
+    return module_name, mod
+
+
+def import_modules(imports, allow_failed_imports=False):
+    """Import modules from the given list of strings.
+
+    Args:
+        imports (list | str | None): The given module names to be imported.
+        allow_failed_imports (bool): If True, the failed imports will return
+            None. Otherwise, an ImportError is raise. Default: False.
+
+    Returns:
+        list[module] | module | None: The imported modules.
+
+    Examples:
+        >>> osp, sys = import_modules(
+        ...     ['os.path', 'sys'])
+        >>> import os.path as osp_
+        >>> import sys as sys_
+        >>> assert osp == osp_
+        >>> assert sys == sys_
+    """
+    if not imports:
+        return
+    single_import = False
+    if isinstance(imports, str):
+        single_import = True
+        imports = [imports]
+    if not isinstance(imports, list):
+        raise TypeError(
+            f'custom_imports must be a list but got type {type(imports)}')
+    imported = []
+    for imp in imports:
+        if not isinstance(imp, str):
+            raise TypeError(
+                f'{imp} is of type {type(imp)} and cannot be imported.')
+        try:
+            imported_tmp = import_module(imp)
+        except ImportError:
+            if allow_failed_imports:
+                logger.warning(f'{imp} failed to import and is ignored.')
+                imported_tmp = None
+            else:
+                raise ImportError
+        imported.append(imported_tmp)
+    if single_import:
+        imported = imported[0]
+    return imported
+
+
+def validate_py_syntax(filename):
+    with open(filename, 'r', encoding='utf-8') as f:
+        # Setting encoding explicitly to resolve coding issue on windows
+        content = f.read()
+    try:
+        ast.parse(content)
+    except SyntaxError as e:
+        raise SyntaxError('There are syntax errors in config '
+                          f'file {filename}: {e}')
+
+
+# following code borrows implementation from huggingface/transformers
+ENV_VARS_TRUE_VALUES = {'1', 'ON', 'YES', 'TRUE'}
+ENV_VARS_TRUE_AND_AUTO_VALUES = ENV_VARS_TRUE_VALUES.union({'AUTO'})
+USE_TF = os.environ.get('USE_TF', 'AUTO').upper()
+USE_TORCH = os.environ.get('USE_TORCH', 'AUTO').upper()
+_torch_version = 'N/A'
+if USE_TORCH in ENV_VARS_TRUE_AND_AUTO_VALUES and USE_TF not in ENV_VARS_TRUE_VALUES:
+    _torch_available = importlib.util.find_spec('torch') is not None
+    if _torch_available:
+        try:
+            _torch_version = importlib_metadata.version('torch')
+            logger.info(f'PyTorch version {_torch_version} available.')
+        except importlib_metadata.PackageNotFoundError:
+            _torch_available = False
+else:
+    logger.info('Disabling PyTorch because USE_TF is set')
+    _torch_available = False
+
+_tf_version = 'N/A'
+if USE_TF in ENV_VARS_TRUE_AND_AUTO_VALUES and USE_TORCH not in ENV_VARS_TRUE_VALUES:
+    _tf_available = importlib.util.find_spec('tensorflow') is not None
+    if _tf_available:
+        candidates = (
+            'tensorflow',
+            'tensorflow-cpu',
+            'tensorflow-gpu',
+            'tf-nightly',
+            'tf-nightly-cpu',
+            'tf-nightly-gpu',
+            'intel-tensorflow',
+            'intel-tensorflow-avx512',
+            'tensorflow-rocm',
+            'tensorflow-macos',
+        )
+        _tf_version = None
+        # For the metadata, we have to look for both tensorflow and tensorflow-cpu
+        for pkg in candidates:
+            try:
+                _tf_version = importlib_metadata.version(pkg)
+                break
+            except importlib_metadata.PackageNotFoundError:
+                pass
+        _tf_available = _tf_version is not None
+    if _tf_available:
+        if version.parse(_tf_version) < version.parse('2'):
+            pass
+        else:
+            logger.info(f'TensorFlow version {_tf_version} available.')
+else:
+    logger.info('Disabling Tensorflow because USE_TORCH is set')
+    _tf_available = False
+
+_timm_available = importlib.util.find_spec('timm') is not None
+try:
+    _timm_version = importlib_metadata.version('timm')
+    logger.debug(f'Successfully imported timm version {_timm_version}')
+except importlib_metadata.PackageNotFoundError:
+    _timm_available = False
+
+
+def is_scipy_available():
+    return importlib.util.find_spec('scipy') is not None
+
+
+def is_sklearn_available():
+    if importlib.util.find_spec('sklearn') is None:
+        return False
+    return is_scipy_available() and importlib.util.find_spec('sklearn.metrics')
+
+
+def is_sentencepiece_available():
+    return importlib.util.find_spec('sentencepiece') is not None
+
+
+def is_protobuf_available():
+    if importlib.util.find_spec('google') is None:
+        return False
+    return importlib.util.find_spec('google.protobuf') is not None
+
+
+def is_tokenizers_available():
+    return importlib.util.find_spec('tokenizers') is not None
+
+
+def is_timm_available():
+    return _timm_available
+
+
+def is_torch_available():
+    return _torch_available
+
+
+def is_torch_cuda_available():
+    if is_torch_available():
+        import torch
+
+        return torch.cuda.is_available()
+    else:
+        return False
+
+
+def is_tf_available():
+    return _tf_available
+
+
+# docstyle-ignore
+PROTOBUF_IMPORT_ERROR = """
+{0} requires the protobuf library but it was not found in your environment. Checkout the instructions on the
+installation page of its repo: https://github.com/protocolbuffers/protobuf/tree/master/python#installation and
+follow the ones that match your environment.
+"""
+
+# docstyle-ignore
+SENTENCEPIECE_IMPORT_ERROR = """
+{0} requires the SentencePiece library but it was not found in your environment. Checkout the instructions on the
+installation page of its repo: https://github.com/google/sentencepiece#installation and follow the ones
+that match your environment.
+"""
+
+# docstyle-ignore
+SKLEARN_IMPORT_ERROR = """
+{0} requires the scikit-learn library but it was not found in your environment. You can install it with:
+```
+pip install -U scikit-learn
+```
+In a notebook or a colab, you can install it by executing a cell with
+```
+!pip install -U scikit-learn
+```
+"""
+
+# docstyle-ignore
+TENSORFLOW_IMPORT_ERROR = """
+{0} requires the TensorFlow library but it was not found in your environment. Checkout the instructions on the
+installation page: https://www.tensorflow.org/install and follow the ones that match your environment.
+"""
+
+# docstyle-ignore
+TIMM_IMPORT_ERROR = """
+{0} requires the timm library but it was not found in your environment. You can install it with pip:
+`pip install timm`
+"""
+
+# docstyle-ignore
+TOKENIZERS_IMPORT_ERROR = """
+{0} requires the 🤗 Tokenizers library but it was not found in your environment. You can install it with:
+```
+pip install tokenizers
+```
+In a notebook or a colab, you can install it by executing a cell with
+```
+!pip install tokenizers
+```
+"""
+
+# docstyle-ignore
+PYTORCH_IMPORT_ERROR = """
+{0} requires the PyTorch library but it was not found in your environment. Checkout the instructions on the
+installation page: https://pytorch.org/get-started/locally/ and follow the ones that match your environment.
+"""
+
+# docstyle-ignore
+SCIPY_IMPORT_ERROR = """
+{0} requires the scipy library but it was not found in your environment. You can install it with pip:
+`pip install scipy`
+"""
+
+REQUIREMENTS_MAAPING = OrderedDict([
+    ('protobuf', (is_protobuf_available, PROTOBUF_IMPORT_ERROR)),
+    ('sentencepiece', (is_sentencepiece_available,
+                       SENTENCEPIECE_IMPORT_ERROR)),
+    ('sklearn', (is_sklearn_available, SKLEARN_IMPORT_ERROR)),
+    ('tf', (is_tf_available, TENSORFLOW_IMPORT_ERROR)),
+    ('timm', (is_timm_available, TIMM_IMPORT_ERROR)),
+    ('tokenizers', (is_tokenizers_available, TOKENIZERS_IMPORT_ERROR)),
+    ('torch', (is_torch_available, PYTORCH_IMPORT_ERROR)),
+    ('scipy', (is_scipy_available, SCIPY_IMPORT_ERROR)),
+])
+
+
+def requires(obj, requirements):
+    if not isinstance(requirements, (list, tuple)):
+        requirements = [requirements]
+    if isinstance(obj, str):
+        name = obj
+    else:
+        name = obj.__name__ if hasattr(obj,
+                                       '__name__') else obj.__class__.__name__
+    checks = (REQUIREMENTS_MAAPING[req] for req in requirements)
+    failed = [msg.format(name) for available, msg in checks if not available()]
+    if failed:
+        raise ImportError(''.join(failed))
+
+
+def torch_required(func):
+    # Chose a different decorator name than in tests so it's clear they are not the same.
+    @functools.wraps(func)
+    def wrapper(*args, **kwargs):
+        if is_torch_available():
+            return func(*args, **kwargs)
+        else:
+            raise ImportError(f'Method `{func.__name__}` requires PyTorch.')
+
+    return wrapper
+
+
+def tf_required(func):
+    # Chose a different decorator name than in tests so it's clear they are not the same.
+    @functools.wraps(func)
+    def wrapper(*args, **kwargs):
+        if is_tf_available():
+            return func(*args, **kwargs)
+        else:
+            raise ImportError(f'Method `{func.__name__}` requires TF.')
+
+    return wrapper

modelscope/utils/pymod.py

@@ -1,90 +0,0 @@
-# Copyright (c) Alibaba, Inc. and its affiliates.
-
-import ast
-import os
-import os.path as osp
-import sys
-import types
-from importlib import import_module
-
-from modelscope.utils.logger import get_logger
-
-logger = get_logger()
-
-
-def import_modules_from_file(py_file: str):
-    """ Import module from a certrain file
-
-    Args:
-        py_file: path to a python file to be imported
-
-    Return:
-
-    """
-    dirname, basefile = os.path.split(py_file)
-    if dirname == '':
-        dirname == './'
-    module_name = osp.splitext(basefile)[0]
-    sys.path.insert(0, dirname)
-    validate_py_syntax(py_file)
-    mod = import_module(module_name)
-    sys.path.pop(0)
-    return module_name, mod
-
-
-def import_modules(imports, allow_failed_imports=False):
-    """Import modules from the given list of strings.
-
-    Args:
-        imports (list | str | None): The given module names to be imported.
-        allow_failed_imports (bool): If True, the failed imports will return
-            None. Otherwise, an ImportError is raise. Default: False.
-
-    Returns:
-        list[module] | module | None: The imported modules.
-
-    Examples:
-        >>> osp, sys = import_modules(
-        ...     ['os.path', 'sys'])
-        >>> import os.path as osp_
-        >>> import sys as sys_
-        >>> assert osp == osp_
-        >>> assert sys == sys_
-    """
-    if not imports:
-        return
-    single_import = False
-    if isinstance(imports, str):
-        single_import = True
-        imports = [imports]
-    if not isinstance(imports, list):
-        raise TypeError(
-            f'custom_imports must be a list but got type {type(imports)}')
-    imported = []
-    for imp in imports:
-        if not isinstance(imp, str):
-            raise TypeError(
-                f'{imp} is of type {type(imp)} and cannot be imported.')
-        try:
-            imported_tmp = import_module(imp)
-        except ImportError:
-            if allow_failed_imports:
-                logger.warning(f'{imp} failed to import and is ignored.')
-                imported_tmp = None
-            else:
-                raise ImportError
-        imported.append(imported_tmp)
-    if single_import:
-        imported = imported[0]
-    return imported
-
-
-def validate_py_syntax(filename):
-    with open(filename, 'r', encoding='utf-8') as f:
-        # Setting encoding explicitly to resolve coding issue on windows
-        content = f.read()
-    try:
-        ast.parse(content)
-    except SyntaxError as e:
-        raise SyntaxError('There are syntax errors in config '
-                          f'file {filename}: {e}')

modelscope/utils/registry.py

@@ -1,7 +1,9 @@
 # Copyright (c) Alibaba, Inc. and its affiliates.
 
 import inspect
+from typing import List, Tuple, Union
 
+from modelscope.utils.import_utils import requires
 from modelscope.utils.logger import get_logger
 
 default_group = 'default'
@@ -52,9 +54,14 @@ class Registry(object):
     def _register_module(self,
                          group_key=default_group,
                          module_name=None,
-                         module_cls=None):
+                         module_cls=None,
+                         requirements=None):
         assert isinstance(group_key,
                           str), 'group_key is required and must be str'
+
+        if requirements is not None:
+            requires(module_cls, requirements)
+
         if group_key not in self._modules:
             self._modules[group_key] = dict()
 
@@ -70,23 +77,11 @@ class Registry(object):
         self._modules[group_key][module_name] = module_cls
         module_cls.group_key = group_key
 
-        if module_name in self._modules[default_group]:
-            if id(self._modules[default_group][module_name]) == id(module_cls):
-                return
-            else:
-                logger.warning(f'{module_name} is already registered in '
-                               f'{self._name}[{default_group}] and will '
-                               'be overwritten')
-                logger.warning(f'{self._modules[default_group][module_name]}'
-                               f'to {module_cls}')
-        # also register module in the default group for faster access
-        # only by module name
-        self._modules[default_group][module_name] = module_cls
-
     def register_module(self,
                         group_key: str = default_group,
                         module_name: str = None,
-                        module_cls: type = None):
+                        module_cls: type = None,
+                        requirements: Union[List, Tuple] = None):
         """ Register module
 
         Example:
@@ -110,17 +105,18 @@ class Registry(object):
                 default group name is 'default'
             module_name: Module name
             module_cls: Module class object
+            requirements: Module necessary requirements
 
         """
         if not (module_name is None or isinstance(module_name, str)):
             raise TypeError(f'module_name must be either of None, str,'
                             f'got {type(module_name)}')
-
         if module_cls is not None:
             self._register_module(
                 group_key=group_key,
                 module_name=module_name,
-                module_cls=module_cls)
+                module_cls=module_cls,
+                requirements=requirements)
             return module_cls
 
         # if module_cls is None, should return a decorator function
@@ -128,7 +124,8 @@ class Registry(object):
             self._register_module(
                 group_key=group_key,
                 module_name=module_name,
-                module_cls=module_cls)
+                module_cls=module_cls,
+                requirements=requirements)
             return module_cls
 
         return _register

requirements.txt

@@ -1,6 +1 @@
 -r requirements/runtime.txt
--r requirements/pipeline.txt
--r requirements/multi-modal.txt
--r requirements/nlp.txt
--r requirements/audio.txt
--r requirements/cv.txt

requirements/audio.txt

@@ -1,10 +1,5 @@
 #tts
 h5py
-https://isv-data.oss-cn-hangzhou.aliyuncs.com/ics/MaaS/TTS/requirements/pytorch_wavelets-1.3.0-py3-none-any.whl
-https://isv-data.oss-cn-hangzhou.aliyuncs.com/ics/MaaS/TTS/requirements/ttsfrd-0.0.2-cp36-cp36m-linux_x86_64.whl; python_version=='3.6'
-https://isv-data.oss-cn-hangzhou.aliyuncs.com/ics/MaaS/TTS/requirements/ttsfrd-0.0.2-cp37-cp37m-linux_x86_64.whl; python_version=='3.7'
-https://isv-data.oss-cn-hangzhou.aliyuncs.com/ics/MaaS/TTS/requirements/ttsfrd-0.0.2-cp38-cp38-linux_x86_64.whl; python_version=='3.8'
-https://isv-data.oss-cn-hangzhou.aliyuncs.com/ics/MaaS/TTS/requirements/ttsfrd-0.0.2-cp39-cp39-linux_x86_64.whl; python_version=='3.9'
 inflect
 keras
 librosa
@@ -14,6 +9,7 @@ nara_wpe
 numpy
 protobuf>3,<=3.20
 ptflops
+pytorch_wavelets==1.3.0
 PyWavelets>=1.0.0
 scikit-learn
 SoundFile>0.10
@@ -24,4 +20,5 @@ torch
 torchaudio
 torchvision
 tqdm
+ttsfrd==0.0.2
 unidecode

requirements/multi-modal.txt

@@ -1,8 +1,6 @@
-datasets
-einops
+fairseq==maas
 ftfy>=6.0.3
-https://jirenmr.oss-cn-zhangjiakou.aliyuncs.com/ofa/fairseq-maas-py3-none-any.whl
-https://jirenmr.oss-cn-zhangjiakou.aliyuncs.com/ofa/ofa-0.0.2-py3-none-any.whl
+ofa==0.0.2
 pycocoevalcap>=1.2
 pycocotools>=2.0.4
 rouge_score

requirements/nlp.txt

@@ -1,3 +1,4 @@
 http://ait-public.oss-cn-hangzhou-zmf.aliyuncs.com/jizhu/en_core_web_sm-2.3.1.tar.gz
 https://alinlp.alibaba-inc.com/pypi/sofa-1.0.5-py3-none-any.whl
+sofa==1.0.5
 spacy>=2.3.5

requirements/pipeline.txt

@@ -1,6 +0,0 @@
-#https://atp-modelzoo-sh.oss-cn-shanghai.aliyuncs.com/release/package/whl/easynlp-0.0.4-py2.py3-none-any.whl
-# tensorflow
-#--find-links https://download.pytorch.org/whl/torch_stable.html
-# torch<1.10,>=1.8.0
-# torchaudio
-# torchvision

requirements/runtime.txt

@@ -1,16 +1,18 @@
 addict
 datasets
 easydict
+einops
 filelock>=3.3.0
 numpy
-opencv-python-headless
+opencv-python
 Pillow>=6.2.0
+protobuf>3,<=3.20
 pyyaml
 requests
-requests==2.27.1
 scipy
-setuptools==58.0.4
+setuptools
 tokenizers<=0.10.3
+torch
 tqdm>=4.64.0
-transformers<=4.16.2
+transformers<=4.16.2,>=4.10.3
 yapf

setup.py

@@ -5,6 +5,8 @@ import shutil
 import subprocess
 from setuptools import find_packages, setup
 
+from modelscope.utils.constant import Fields
+
 
 def readme():
     with open('README.md', encoding='utf-8') as f:
@@ -169,6 +171,16 @@ if __name__ == '__main__':
     pack_resource()
     os.chdir('package')
     install_requires, deps_link = parse_requirements('requirements.txt')
+    extra_requires = {}
+    all_requires = []
+    for field in dir(Fields):
+        if field.startswith('_'):
+            continue
+        extra_requires[field], _ = parse_requirements(
+            f'requirements/{field}.txt')
+        all_requires.append(extra_requires[field])
+    extra_requires['all'] = all_requires
+
     setup(
         name='model-scope',
         version=get_version(),
@@ -193,5 +205,6 @@ if __name__ == '__main__':
         license='Apache License 2.0',
         tests_require=parse_requirements('requirements/tests.txt'),
         install_requires=install_requires,
+        extras_require=extra_requires,
         dependency_links=deps_link,
         zip_safe=False)

tests/pipelines/test_base.py

@@ -8,6 +8,7 @@ import PIL
 
 from modelscope.pipelines import Pipeline, pipeline
 from modelscope.pipelines.builder import PIPELINES, add_default_pipeline_info
+from modelscope.pipelines.outputs import OutputKeys
 from modelscope.utils.constant import Tasks
 from modelscope.utils.logger import get_logger
 from modelscope.utils.registry import default_group
@@ -68,28 +69,28 @@ class CustomPipelineTest(unittest.TestCase):
                     outputs['filename'] = inputs['url']
                 img = inputs['img']
                 new_image = img.resize((img.width // 2, img.height // 2))
-                outputs['output_png'] = np.array(new_image)
+                outputs[OutputKeys.OUTPUT_IMG] = np.array(new_image)
                 return outputs
 
             def postprocess(self, inputs: Dict[str, Any]) -> Dict[str, Any]:
                 return inputs
 
-        self.assertTrue('custom-image' in PIPELINES.modules[default_group])
+        self.assertTrue('custom-image' in PIPELINES.modules[dummy_task])
         add_default_pipeline_info(dummy_task, 'custom-image', overwrite=True)
-        pipe = pipeline(pipeline_name='custom-image')
+        pipe = pipeline(task=dummy_task, pipeline_name='custom-image')
         pipe2 = pipeline(dummy_task)
         self.assertTrue(type(pipe) is type(pipe2))
 
         img_url = 'data/test/images/image1.jpg'
         output = pipe(img_url)
         self.assertEqual(output['filename'], img_url)
-        self.assertEqual(output['output_png'].shape, (318, 512, 3))
+        self.assertEqual(output[OutputKeys.OUTPUT_IMG].shape, (318, 512, 3))
 
         outputs = pipe([img_url for i in range(4)])
         self.assertEqual(len(outputs), 4)
         for out in outputs:
             self.assertEqual(out['filename'], img_url)
-            self.assertEqual(out['output_png'].shape, (318, 512, 3))
+            self.assertEqual(out[OutputKeys.OUTPUT_IMG].shape, (318, 512, 3))
 
 
 if __name__ == '__main__':

tests/pipelines/test_image_captioning.py

@@ -3,6 +3,7 @@
 import unittest
 
 from modelscope.pipelines import pipeline
+from modelscope.pipelines.outputs import OutputKeys
 from modelscope.utils.constant import Tasks
 from modelscope.utils.test_utils import test_level
 
@@ -15,7 +16,7 @@ class ImageCaptionTest(unittest.TestCase):
             Tasks.image_captioning,
             model='damo/ofa_image-caption_coco_large_en')
         result = img_captioning('data/test/images/image_captioning.png')
-        print(result['caption'])
+        print(result[OutputKeys.CAPTION])
 
 
 if __name__ == '__main__':

tests/pipelines/test_image_matting.py

@@ -9,6 +9,7 @@ import cv2
 from modelscope.fileio import File
 from modelscope.msdatasets import MsDataset
 from modelscope.pipelines import pipeline
+from modelscope.pipelines.outputs import OutputKeys
 from modelscope.utils.constant import ModelFile, Tasks
 from modelscope.utils.test_utils import test_level
 
@@ -29,7 +30,7 @@ class ImageMattingTest(unittest.TestCase):
             img_matting = pipeline(Tasks.image_matting, model=tmp_dir)
 
             result = img_matting('data/test/images/image_matting.png')
-            cv2.imwrite('result.png', result['output_png'])
+            cv2.imwrite('result.png', result[OutputKeys.OUTPUT_IMG])
 
     @unittest.skipUnless(test_level() >= 1, 'skip test in current test level')
     def test_run_with_dataset(self):
@@ -41,7 +42,7 @@ class ImageMattingTest(unittest.TestCase):
         img_matting = pipeline(Tasks.image_matting, model=self.model_id)
         # note that for dataset output, the inference-output is a Generator that can be iterated.
         result = img_matting(dataset)
-        cv2.imwrite('result.png', next(result)['output_png'])
+        cv2.imwrite('result.png', next(result)[OutputKeys.OUTPUT_IMG])
         print(f'Output written to {osp.abspath("result.png")}')
 
     @unittest.skipUnless(test_level() >= 0, 'skip test in current test level')
@@ -49,7 +50,7 @@ class ImageMattingTest(unittest.TestCase):
         img_matting = pipeline(Tasks.image_matting, model=self.model_id)
 
         result = img_matting('data/test/images/image_matting.png')
-        cv2.imwrite('result.png', result['output_png'])
+        cv2.imwrite('result.png', result[OutputKeys.OUTPUT_IMG])
         print(f'Output written to {osp.abspath("result.png")}')
 
     @unittest.skipUnless(test_level() >= 2, 'skip test in current test level')
@@ -57,7 +58,7 @@ class ImageMattingTest(unittest.TestCase):
         img_matting = pipeline(Tasks.image_matting)
 
         result = img_matting('data/test/images/image_matting.png')
-        cv2.imwrite('result.png', result['output_png'])
+        cv2.imwrite('result.png', result[OutputKeys.OUTPUT_IMG])
         print(f'Output written to {osp.abspath("result.png")}')
 
     @unittest.skipUnless(test_level() >= 2, 'skip test in current test level')
@@ -67,7 +68,7 @@ class ImageMattingTest(unittest.TestCase):
         img_matting = pipeline(Tasks.image_matting, model=self.model_id)
         result = img_matting(dataset)
         for i in range(10):
-            cv2.imwrite(f'result_{i}.png', next(result)['output_png'])
+            cv2.imwrite(f'result_{i}.png', next(result)[OutputKeys.OUTPUT_IMG])
         print(
             f'Output written to dir: {osp.dirname(osp.abspath("result_0.png"))}'
         )

tests/pipelines/test_key_word_spotting.py

@@ -15,8 +15,8 @@ from modelscope.utils.test_utils import test_level
 
 KWSBP_URL = 'https://isv-data.oss-cn-hangzhou.aliyuncs.com/ics/MaaS/KWS/tools/kwsbp'
 
-POS_WAV_FILE = '20200707_spk57db_storenoise52db_40cm_xiaoyun_sox_6.wav'
-POS_WAV_URL = 'https://isv-data.oss-cn-hangzhou.aliyuncs.com/ics/MaaS/KWS/pos_testset/' + POS_WAV_FILE
+POS_WAV_FILE = 'data/test/audios/kws_xiaoyunxiaoyun.wav'
+BOFANGYINYUE_WAV_FILE = 'data/test/audios/kws_bofangyinyue.wav'
 
 POS_TESTSETS_FILE = 'pos_testsets.tar.gz'
 POS_TESTSETS_URL = 'https://isv-data.oss-cn-hangzhou.aliyuncs.com/ics/MaaS/KWS/pos_testsets.tar.gz'
@@ -47,12 +47,8 @@ class KeyWordSpottingTest(unittest.TestCase):
         # wav, neg_testsets, pos_testsets, roc
         kws_set = 'wav'
 
-        # downloading wav file
-        wav_file_path = os.path.join(self.workspace, POS_WAV_FILE)
-        if not os.path.exists(wav_file_path):
-            r = requests.get(POS_WAV_URL)
-            with open(wav_file_path, 'wb') as f:
-                f.write(r.content)
+        # get wav file
+        wav_file_path = POS_WAV_FILE
 
         # downloading kwsbp
         kwsbp_file_path = os.path.join(self.workspace, 'kwsbp')
@@ -70,6 +66,7 @@ class KeyWordSpottingTest(unittest.TestCase):
         self.assertTrue(preprocessor is not None)
 
         kwsbp_16k_pipline = pipeline(
+            task=Tasks.key_word_spotting,
             pipeline_name=Pipelines.kws_kwsbp,
             model=model,
             preprocessor=preprocessor)
@@ -91,9 +88,73 @@ class KeyWordSpottingTest(unittest.TestCase):
         """
         if kws_result.__contains__('keywords'):
             print('test_run_with_wav keywords: ', kws_result['keywords'])
+            print('test_run_with_wav confidence: ', kws_result['confidence'])
         print('test_run_with_wav detected result: ', kws_result['detected'])
         print('test_run_with_wav wave time(seconds): ', kws_result['wav_time'])
 
+    @unittest.skipUnless(test_level() >= 0, 'skip test in current test level')
+    def test_run_with_wav_by_customized_keywords(self):
+        # wav, neg_testsets, pos_testsets, roc
+        kws_set = 'wav'
+
+        # get wav file
+        wav_file_path = BOFANGYINYUE_WAV_FILE
+
+        # downloading kwsbp
+        kwsbp_file_path = os.path.join(self.workspace, 'kwsbp')
+        if not os.path.exists(kwsbp_file_path):
+            r = requests.get(KWSBP_URL)
+            with open(kwsbp_file_path, 'wb') as f:
+                f.write(r.content)
+
+        model = Model.from_pretrained(self.model_id)
+        self.assertTrue(model is not None)
+
+        cfg_preprocessor = dict(
+            type=Preprocessors.wav_to_lists, workspace=self.workspace)
+        preprocessor = build_preprocessor(cfg_preprocessor, Fields.audio)
+        self.assertTrue(preprocessor is not None)
+
+        # customized keyword if you need.
+        # full settings eg.
+        #     keywords = [
+        #         {'keyword':'你好电视', 'threshold': 0.008},
+        #         {'keyword':'播放音乐', 'threshold': 0.008}
+        #     ]
+        keywords = [{'keyword': '播放音乐'}]
+
+        kwsbp_16k_pipline = pipeline(
+            task=Tasks.key_word_spotting,
+            pipeline_name=Pipelines.kws_kwsbp,
+            model=model,
+            preprocessor=preprocessor,
+            keywords=keywords)
+        self.assertTrue(kwsbp_16k_pipline is not None)
+
+        kws_result = kwsbp_16k_pipline(
+            kws_type=kws_set, wav_path=[wav_file_path, None])
+        self.assertTrue(kws_result.__contains__('detected'))
+        """
+        kws result json format example:
+            {
+                'wav_count': 1,
+                'kws_set': 'wav',
+                'wav_time': 9.132938,
+                'keywords': ['播放音乐'],
+                'detected': True,
+                'confidence': 0.660368
+            }
+        """
+        if kws_result.__contains__('keywords'):
+            print('test_run_with_wav_by_customized_keywords keywords: ',
+                  kws_result['keywords'])
+            print('test_run_with_wav_by_customized_keywords confidence: ',
+                  kws_result['confidence'])
+        print('test_run_with_wav_by_customized_keywords detected result: ',
+              kws_result['detected'])
+        print('test_run_with_wav_by_customized_keywords wave time(seconds): ',
+              kws_result['wav_time'])
+
     @unittest.skipUnless(test_level() >= 1, 'skip test in current test level')
     def test_run_with_pos_testsets(self):
         # wav, neg_testsets, pos_testsets, roc
@@ -133,6 +194,7 @@ class KeyWordSpottingTest(unittest.TestCase):
         self.assertTrue(preprocessor is not None)
 
         kwsbp_16k_pipline = pipeline(
+            task=Tasks.key_word_spotting,
             pipeline_name=Pipelines.kws_kwsbp,
             model=model,
             preprocessor=preprocessor)
@@ -204,6 +266,7 @@ class KeyWordSpottingTest(unittest.TestCase):
         self.assertTrue(preprocessor is not None)
 
         kwsbp_16k_pipline = pipeline(
+            task=Tasks.key_word_spotting,
             pipeline_name=Pipelines.kws_kwsbp,
             model=model,
             preprocessor=preprocessor)
@@ -298,6 +361,7 @@ class KeyWordSpottingTest(unittest.TestCase):
         self.assertTrue(preprocessor is not None)
 
         kwsbp_16k_pipline = pipeline(
+            task=Tasks.key_word_spotting,
             pipeline_name=Pipelines.kws_kwsbp,
             model=model,
             preprocessor=preprocessor)

tests/pipelines/test_person_image_cartoon.py

@@ -7,6 +7,7 @@ import cv2
 
 from modelscope.pipelines import pipeline
 from modelscope.pipelines.base import Pipeline
+from modelscope.pipelines.outputs import OutputKeys
 from modelscope.utils.constant import Tasks
 from modelscope.utils.test_utils import test_level
 
@@ -22,7 +23,7 @@ class ImageCartoonTest(unittest.TestCase):
     def pipeline_inference(self, pipeline: Pipeline, input_location: str):
         result = pipeline(input_location)
         if result is not None:
-            cv2.imwrite('result.png', result['output_png'])
+            cv2.imwrite('result.png', result[OutputKeys.OUTPUT_IMG])
             print(f'Output written to {osp.abspath("result.png")}')
 
     @unittest.skip('deprecated, download model from model hub instead')

tests/pipelines/test_text_to_speech.py

@@ -12,7 +12,7 @@ from modelscope.metainfo import Pipelines, Preprocessors
 from modelscope.models import Model
 from modelscope.pipelines import pipeline
 from modelscope.preprocessors import build_preprocessor
-from modelscope.utils.constant import Fields
+from modelscope.utils.constant import Fields, Tasks
 from modelscope.utils.logger import get_logger
 from modelscope.utils.test_utils import test_level
 
@@ -43,6 +43,7 @@ class TextToSpeechSambertHifigan16kPipelineTest(unittest.TestCase):
         self.assertTrue(voc is not None)
 
         sambert_tts = pipeline(
+            task=Tasks.text_to_speech,
             pipeline_name=Pipelines.sambert_hifigan_16k_tts,
             config_file='',
             model=[am, voc],

tests/utils/test_check_requirements.py

@@ -0,0 +1,22 @@
+# Copyright (c) Alibaba, Inc. and its affiliates.
+
+import unittest
+from typing import List, Union
+
+from modelscope.utils.check_requirements import NLPModuleNotFoundError, get_msg
+from modelscope.utils.constant import Fields
+
+
+class ImportUtilsTest(unittest.TestCase):
+
+    def test_type_module_not_found(self):
+        with self.assertRaises(NLPModuleNotFoundError) as ctx:
+            try:
+                import not_found
+            except ModuleNotFoundError as e:
+                raise NLPModuleNotFoundError(e)
+        self.assertTrue(get_msg(Fields.nlp) in ctx.exception.msg.msg)
+
+
+if __name__ == '__main__':
+    unittest.main()