refine tests and examples

Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/8898823

docs/source/quick_start.md

@@ -20,7 +20,7 @@ which pip
 
 ## 第三方依赖安装
 
-MaaS Library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:
+MaaS 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)
@@ -41,7 +41,7 @@ python -c "from maas_lib.pipelines import pipeline;print(pipeline('image-matting
 ```
 
 
-### 使用源码
+### 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
 ```shell
@@ -64,7 +64,6 @@ python -c "from maas_lib.pipelines import pipeline;print(pipeline('image-matting
 ```
 
 
-
 ## 训练
 
 to be done
@@ -84,12 +83,33 @@ from maas_lib.pipelines import pipeline
 from maas_lib.utils.constant import Tasks
 
 # 根据任务名创建pipeline
-img_matting = pipeline(
-    Tasks.image_matting, model='damo/image-matting-person')
+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'result file path is {osp.abspath("result.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 maas_lib.pipelines import pipeline
+from maas_lib.utils.constant import Tasks
+from ali_maas_datasets import PyDataset
+
+# 使用图像url构建PyDataset，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 = PyDataset.load(input_location, target='image')
+img_matting = pipeline(Tasks.image_matting, model='damo/image-matting-person')
+# 输入为PyDataset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")}')
 ```

tests/pipelines/test_image_matting.py

@@ -3,11 +3,8 @@
 import os.path as osp
 import tempfile
 import unittest
-from typing import Any, Dict, List, Tuple, Union
 
 import cv2
-import numpy as np
-import PIL
 from ali_maas_datasets import PyDataset
 
 from maas_lib.fileio import File
@@ -31,24 +28,20 @@ class ImageMattingTest(unittest.TestCase):
             )
             cv2.imwrite('result.png', result['output_png'])
 
-    def test_dataset(self):
-        model_path = 'http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs' \
-                     '.com/data/test/maas/image_matting/matting_person.pb'
-        with tempfile.TemporaryDirectory() as tmp_dir:
-            model_file = osp.join(tmp_dir, 'matting_person.pb')
-            with open(model_file, 'wb') as ofile:
-                ofile.write(File.read(model_path))
-            img_matting = pipeline(Tasks.image_matting, model=tmp_dir)
-            # dataset = PyDataset.load('/dir/to/images', target='image')
-            # yapf: disable
-            dataset = PyDataset.load([
-                'http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/data/test/maas/image_matting/test.png'
-            ],
-                target='image')
-            result = img_matting(dataset)
-            for i, r in enumerate(result):
-                cv2.imwrite(f'/path/to/result/{i}.png', r['output_png'])
-            print('end')
+    def test_run_with_dataset(self):
+        input_location = [
+            'http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/data/test/maas/image_matting/test.png'
+        ]
+        # alternatively:
+        # input_location = '/dir/to/images'
+
+        dataset = PyDataset.load(input_location, target='image')
+        img_matting = pipeline(
+            Tasks.image_matting, model='damo/image-matting-person')
+        # 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'])
+        print(f'Output written to {osp.abspath("result.png")}')
 
     def test_run_modelhub(self):
         img_matting = pipeline(
@@ -58,6 +51,7 @@ class ImageMattingTest(unittest.TestCase):
             '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")}')
 
 
 if __name__ == '__main__':

tests/pipelines/test_text_classification.py

@@ -11,6 +11,7 @@ from maas_lib.models import Model
 from maas_lib.models.nlp import SequenceClassificationModel
 from maas_lib.pipelines import SequenceClassificationPipeline, pipeline
 from maas_lib.preprocessors import SequenceClassificationPreprocessor
+from maas_lib.utils.constant import Tasks
 
 
 class SequenceClassificationTest(unittest.TestCase):
@@ -49,7 +50,7 @@ class SequenceClassificationTest(unittest.TestCase):
         pipeline1 = SequenceClassificationPipeline(model, preprocessor)
         self.predict(pipeline1)
         pipeline2 = pipeline(
-            'text-classification', model=model, preprocessor=preprocessor)
+            Tasks.text_classification, model=model, preprocessor=preprocessor)
         print(pipeline2('Hello world!'))
 
     def test_run_modelhub(self):
@@ -57,29 +58,20 @@ class SequenceClassificationTest(unittest.TestCase):
         preprocessor = SequenceClassificationPreprocessor(
             model.model_dir, first_sequence='sentence', second_sequence=None)
         pipeline_ins = pipeline(
-            task='text-classification', model=model, preprocessor=preprocessor)
+            task=Tasks.text_classification,
+            model=model,
+            preprocessor=preprocessor)
         self.predict(pipeline_ins)
 
-    def test_dataset(self):
-        model_url = 'https://atp-modelzoo-sh.oss-cn-shanghai.aliyuncs.com' \
-                    '/release/easynlp_modelzoo/alibaba-pai/bert-base-sst2.zip'
-        cache_path_str = r'.cache/easynlp/bert-base-sst2.zip'
-        cache_path = Path(cache_path_str)
-
-        if not cache_path.exists():
-            cache_path.parent.mkdir(parents=True, exist_ok=True)
-            cache_path.touch(exist_ok=True)
-            with cache_path.open('wb') as ofile:
-                ofile.write(File.read(model_url))
-
-        with zipfile.ZipFile(cache_path_str, 'r') as zipf:
-            zipf.extractall(cache_path.parent)
-        path = r'.cache/easynlp/'
-        model = SequenceClassificationModel(path)
+    def test_run_with_dataset(self):
+        model = Model.from_pretrained('damo/bert-base-sst2')
         preprocessor = SequenceClassificationPreprocessor(
-            path, first_sequence='sentence', second_sequence=None)
+            model.model_dir, first_sequence='sentence', second_sequence=None)
         text_classification = pipeline(
-            'text-classification', model=model, preprocessor=preprocessor)
+            Tasks.text_classification, model=model, preprocessor=preprocessor)
+        # loaded from huggingface dataset
+        # TODO: add load_from parameter (an enum) LOAD_FROM.hugging_face
+        # TODO: rename parameter as dataset_name and subset_name
         dataset = PyDataset.load('glue', name='sst2', target='sentence')
         result = text_classification(dataset)
         for i, r in enumerate(result):