Merge pull request #163 from SamitHuang/main

rm redundant packages and add ut to workflow

Author: SamitHuang

.github/workflows/ci.yml

@@ -36,10 +36,10 @@ jobs:
         pip install pytest
         # MindSpore must be installed following the instruction from official web, but not from pypi.
         # That's why we exclude mindspore from requirements.txt. Does this work?
-        pip install "mindspore>=1.8,<=1.10"
-    #- name: Test with pytest (UT)
-    #  run: |
-    #    pytest tests/modules/*.py
+        pip install "mindspore>=1.9,<=1.10"
+    - name: Test with pytest (UT)
+      run: |
+        pytest tests/ut/*.py
     - name: Test with pytest (ST)
       run: |
         pytest tests/st/test_train_eval_dummy.py

tests/st/test_train_eval_dummy.py

@@ -12,95 +12,29 @@
 import sys
 import glob
 import yaml
+import pytest
 
 sys.path.append(".")
 
-import pytest
-
+from tests.ut._common import gen_dummpy_data, update_config_for_CI
 from mindocr.models.backbones.mindcv_models.download import DownLoad
 
 
 @pytest.mark.parametrize("task", ["det", "rec"])
 @pytest.mark.parametrize("val_while_train", [False, True])
 def test_train_eval(task, val_while_train):
+
     # prepare dummy images
-    data_dir = "data/Canidae"
-    dataset_url = (
-        "https://mindspore-website.obs.cn-north-4.myhuaweicloud.com/notebook/datasets/intermediate/Canidae_data.zip"
-    )
-    if not os.path.exists(data_dir):
-        DownLoad().download_and_extract_archive(dataset_url, "./")
-
-    # prepare dummy labels
-    for split in ['train', 'val']:
-        label_path = f'tests/st/dummy_labels/{task}_{split}_gt.txt'
-        image_dir = f'{data_dir}/{split}/dogs'
-        new_label_path = f'data/Canidae/{split}/{task}_gt.txt'
-        img_paths = glob.glob(os.path.join(image_dir, '*.JPEG'))
-        #print(len(img_paths))
-        with open(new_label_path, 'w') as f_w:
-            with open(label_path, 'r') as f_r:
-                i = 0
-                for line in f_r:
-                    _, label = line.strip().split('\t')
-                    #print(i)
-                    img_name = os.path.basename(img_paths[i])
-                    new_img_label = img_name + '\t' + label
-                    f_w.write(new_img_label + '\n')
-                    i += 1
-        print(f'Dummpy annotation file is generated in {new_label_path}')
-
-    # modify ocr predefined yaml for minimum test
+    data_dir = gen_dummpy_data(task)
+
+     # modify ocr predefined yaml for minimum test
     if task == 'det':
         config_fp = 'configs/det/dbnet/db_r50_icdar15.yaml'
     elif task=='rec':
         #config_fp = 'configs/rec/vgg7_bilstm_ctc.yaml' # TODO: change on lmdb datasset
         config_fp = 'configs/rec/crnn/crnn_icdar15.yaml'
 
-    with open(config_fp) as fp:
-        config = yaml.safe_load(fp)
-        config['system']['distribute'] = False
-        config['system']['val_while_train'] = val_while_train
-        #if 'common' in config:
-        #    config['batch_size'] = 8
-        config['train']['dataset_sink_mode'] = False
-
-        config['train']['dataset']['dataset_root'] = 'data/Canidae/'
-        config['train']['dataset']['data_dir'] = 'train/dogs'
-        config['train']['dataset']['label_file'] = f'train/{task}_gt.txt'
-        config['train']['dataset']['sample_ratio'] = 0.1 # TODO: 120 training samples in total, don't be larger than batchsize after sampling
-        config['train']['loader']['num_workers'] = 1 # github server only support 2 workers at most
-        #if config['train']['loader']['batch_size'] > 120:
-        config['train']['loader']['batch_size'] = 2 # to save memory
-        config['train']['loader']['max_rowsize'] = 16 # to save memory
-        config['train']['loader']['prefetch_size'] = 2 # to save memory
-        if 'common' in config:
-            config['common']['batch_size'] = 2
-        if 'batch_size' in config['loss']:
-            config['loss']['batch_size'] = 2
-
-        config['eval']['dataset']['dataset_root'] = 'data/Canidae/'
-        config['eval']['dataset']['data_dir'] = 'val/dogs'
-        config['eval']['dataset']['label_file'] = f'val/{task}_gt.txt'
-        config['eval']['dataset']['sample_ratio'] = 0.1
-        config['eval']['loader']['num_workers'] = 1 # github server only support 2 workers at most
-        config['eval']['loader']['batch_size'] = 1
-        config['eval']['loader']['max_rowsize'] = 16 # to save memory
-        config['eval']['loader']['prefetch_size'] = 2 # to save memory
-
-        config['eval']['ckpt_load_path'] = os.path.join(config['train']['ckpt_save_dir'], 'best.ckpt')
-
-        config['scheduler']['num_epochs'] = 2
-        config['scheduler']['warmup_epochs'] = 1
-        config['scheduler']['decay_epochs'] = 1
-
-        dummpy_config_fp =os.path.join('tests/st', os.path.basename(config_fp.replace('.yaml', '_dummpy.yaml')))
-        with open(dummpy_config_fp, 'w') as f:
-            args_text = yaml.safe_dump(config, default_flow_style=False, sort_keys=False)
-            f.write(args_text)
-            print('Genearted yaml: ')
-            print(args_text)
-
+    dummpy_config_fp = update_config_for_CI(config_fp, task)
 
     #dummpy_config_fp = 'tests/st/rec_crnn_test.yaml'
     # ---------------- test running train.py using the toy data ---------
@@ -129,5 +63,5 @@ def test_train_eval(task, val_while_train):
 
 
 if __name__ == '__main__':
-    test_train_eval('det', True)
-    #test_train_eval('rec', True)
+    #test_train_eval('det', True)
+    test_train_eval('rec', True)

tests/ut/_common.py

@@ -1,3 +1,5 @@
+import sys
+sys.path.append('.')
 import os
 import glob
 import yaml

tests/ut/test_datasets.py

@@ -4,19 +4,13 @@
 sys.path.append('.')
 
 import yaml
-import glob
 import pytest
-import numpy as np
 import time
 
 import mindspore as ms
 import mindocr
 from mindocr.data import build_dataset
-from mindocr.data.det_dataset import DetDataset
-from mindocr.data.transforms.transforms_factory import transforms_dbnet_icdar15
-from mindocr.data.rec_dataset import RecDataset
-from mindspore import load_checkpoint, load_param_into_net
-from mindocr.utils.visualize import show_img, draw_bboxes, show_imgs, recover_image
+from mindocr.utils.visualize import show_img, draw_bboxes, recover_image
 
 
 @pytest.mark.parametrize('task', ['det', 'rec'])

tests/ut/test_loss.py

@@ -2,10 +2,8 @@
 sys.path.append('.')
 import pytest
 import yaml
-import numpy as np
-import mindspore as ms
 from addict import Dict
-from mindocr.losses import build_loss
+from mindocr import build_loss
 
 
 @pytest.mark.parametrize('task', ['det', 'rec'])

tests/ut/test_metrics.py

@@ -1,12 +1,30 @@
 import sys
 sys.path.append('.')
-
+import pytest
+import yaml
+from addict import Dict
 import numpy as np
 import mindspore as ms
+from mindocr import build_metric
 from mindocr.metrics.det_metrics import DetMetric
 from mindocr.metrics.rec_metrics import RecMetric
 
 
+@pytest.mark.parametrize('task', ['det', 'rec'])
+@pytest.mark.parametrize('device_num', [1, 8])
+def test_build_metric(task, device_num):
+  if task == 'det':
+      config_fp = 'configs/det/dbnet/db_r50_icdar15.yaml'
+  elif task=='rec':
+      config_fp = 'configs/rec/crnn/crnn_icdar15.yaml'
+
+  with open(config_fp) as fp:
+      cfg = yaml.safe_load(fp)
+  cfg = Dict(cfg)
+
+  metric = build_metric(cfg.metric, device_num=device_num)
+
+
 def test_det_metric():
     pred_polys = [ 
                   [

tests/ut/test_models.py

@@ -1,15 +1,13 @@
 import sys
 sys.path.append('.')
 
+import time
 import yaml
-import glob
 import pytest
 import numpy as np
 import mindspore as ms
 import mindocr
-from mindocr.models.backbones import build_backbone
 from mindocr.models import build_model
-from mindspore import load_checkpoint, load_param_into_net
 
 all_model_names = mindocr.list_models()
 print('Registered models: ', all_model_names)
@@ -20,10 +18,6 @@
 print('All config yamls: ', all_yamls)
 
 def _infer_dummy(model, task='det', verbose=True):
-    import mindspore as ms
-    import time
-    import numpy as np
-
     print(task)
 
     bs = 8
@@ -93,18 +87,5 @@ def test_model_by_yaml(yaml_fp):
 
 if __name__ == '__main__':
     print(all_model_names)
-    #test_model_by_name(all_model_names[0], True)
-    #test_model_by_name(all_model_names[1], True)
     test_model_by_name(all_model_names[2], True)
-    #test_model_by_yaml(all_yamls[1])
-    '''
-    import argparse
-    parser = argparse.ArgumentParser(description='model config', add_help=False)
-    parser.add_argument('-c', '--config', type=str, default='configs/det/dbnet/db_r50_icdar15.yaml',
-                               help='YAML config file specifying default arguments (default='')')
-    args = parser.parse_args()
-    #test_registry()
-    #test_backbone()
-    #test_model_by_name('dbnet_r50')
-    test_model_by_yaml(args.config)
-    '''
+    #test_model_by_yaml(all_yamls[1])

tests/ut/test_postprocess.py

@@ -1,7 +1,6 @@
 import sys
 sys.path.append('.')
 import pytest
-import numpy as np
 import yaml
 from addict import Dict
 from mindocr.postprocess import build_postprocess