Fix pylint issues

Author: sovrasov

.pylintrc

@@ -28,7 +28,7 @@ disable = arguments-differ,
 
 max-line-length = 120
 ignore-docstrings = yes
-ignored-modules = numpy,torch,cv2,openvino
+ignored-modules = numpy,torch,cv2,openvino,onnx.onnx_cpp2py_export
 extension-pkg-whitelist = torch,cv2,scipy
 
 [SIMILARITIES]

requirements.txt

@@ -27,4 +27,4 @@ optuna
 timm
 addict
 randaugment
-pylint==2.6.0
+pylint==2.11.1

scripts/default_config.py

@@ -3,6 +3,7 @@
 
 from yacs.config import CfgNode as CN
 
+# pylint: disable=protected-access,too-many-statements,unspecified-encoding
 
 def get_default_config():
 
@@ -109,7 +110,8 @@ def get_default_config():
 
     # custom_datasets
     cfg.custom_datasets = CN() # this node contains information about custom classification datasets
-    cfg.custom_datasets.roots = [] # a list of root folders in case of ImagesFolder fromat or list of annotation files with paths relative to the list's parent folder
+    cfg.custom_datasets.roots = [] # a list of root folders in case of ImagesFolder fromat
+    # or list of annotation files with paths relative to the list's parent folder
     cfg.custom_datasets.types = [] # a list of types (classification or classification_image_folder)
     cfg.custom_datasets.names = [] # aliases for custom datasets that can be used in the data section. Should be unique
 
@@ -136,7 +138,8 @@ def get_default_config():
     cfg.train.new_layers = ['classifier']  # newly added layers with default lr
     cfg.train.base_lr_mult = 0.1  # learning rate multiplier for base layers
     cfg.train.lr_scheduler = 'single_step'
-    cfg.train.target_metric = 'train_loss' # define which metric to use with reduce_on_plateau scheduler. Two possible variants are available: 'test_acc' and 'train_loss'
+    cfg.train.target_metric = 'train_loss' # define which metric to use with reduce_on_plateau scheduler.
+    # Two possible variants are available: 'test_acc' and 'train_loss'
     cfg.train.base_scheduler = ''
     cfg.train.stepsize = [20]  # stepsize to decay learning rate
     cfg.train.gamma = 0.1  # learning rate decay multiplier
@@ -464,7 +467,7 @@ def get_default_config():
 def merge_from_files_with_base(cfg, cfg_path):
     def _get_list_of_files(cur_path, set_of_files=None):
         if not (cur_path.lower().endswith('.yml') or cur_path.lower().endswith('.yaml')):
-                raise RuntimeError(f'Wrong extension of config file {cur_path}')
+            raise RuntimeError(f'Wrong extension of config file {cur_path}')
         if set_of_files is None:
             set_of_files = {cur_path}
         elif cur_path in set_of_files:

scripts/script_utils.py

@@ -13,6 +13,7 @@
                                     model_kwargs, optimizer_kwargs,
                                     lr_scheduler_kwargs, merge_from_files_with_base)
 
+
 def build_base_argparser():
     parser = argparse.ArgumentParser(formatter_class=argparse.ArgumentDefaultsHelpFormatter)
     parser.add_argument('--config-file', type=str, default='',
@@ -61,6 +62,7 @@ def reset_config(cfg, args):
 def build_datamanager(cfg, classification_classes_filter=None):
     return torchreid.data.ImageDataManager(filter_classes=classification_classes_filter, **imagedata_kwargs(cfg))
 
+
 def build_auxiliary_model(config_file, num_classes, use_gpu,
                           device_ids, num_iter, lr=None,
                           nncf_aux_config_file=None,
@@ -76,14 +78,16 @@ def build_auxiliary_model(config_file, num_classes, use_gpu,
                 f'the changes are:\n{pformat(aux_config_opts)}')
         aux_cfg.merge_from_list(aux_config_opts)
 
-    print('\nShow auxiliary configuration\n{}\n'.format(aux_cfg))
+    print(f'\nShow auxiliary configuration\n{aux_cfg}\n')
 
     if lr is not None:
         aux_cfg.train.lr = lr
         print(f"setting learning rate from main model: {lr}")
     model = torchreid.models.build_model(**model_kwargs(aux_cfg, num_classes))
     optimizer = torchreid.optim.build_optimizer(model, **optimizer_kwargs(aux_cfg))
-    scheduler = torchreid.optim.build_lr_scheduler(optimizer=optimizer, num_iter=num_iter, **lr_scheduler_kwargs(aux_cfg))
+    scheduler = torchreid.optim.build_lr_scheduler(optimizer=optimizer,
+                                                   num_iter=num_iter,
+                                                   **lr_scheduler_kwargs(aux_cfg))
 
     if aux_cfg.model.resume and check_isfile(aux_cfg.model.resume):
         aux_cfg.train.start_epoch = resume_from_checkpoint(
@@ -132,7 +136,7 @@ def is_config_parameter_set_from_command_line(cmd_line_opts, parameter_name):
     if not cmd_line_opts:
         return False
     key_names = cmd_line_opts[0::2]
-    return (parameter_name in key_names)
+    return parameter_name in key_names
 
 
 def put_main_model_on_the_device(model, use_gpu=True, gpu_num=1, num_aux_models=0, split_models=False):
@@ -173,29 +177,30 @@ def check_classes_consistency(ref_classes, probe_classes, strict=False):
             if len(ref_classes) != len(probe_classes):
                 return False
             return sorted(probe_classes.keys()) == sorted(ref_classes.keys())
-        else:
-            if len(ref_classes) > len(probe_classes):
+
+        if len(ref_classes) > len(probe_classes):
+            return False
+        probe_names = probe_classes.keys()
+        for cl in ref_classes.keys():
+            if cl not in probe_names:
                 return False
-            probe_names = probe_classes.keys()
-            for cl in ref_classes.keys():
-                if cl not in probe_names:
-                    return False
+
         return True
 
     classes_map = {v : k for k, v in enumerate(sorted(classes))} if classes else {}
     if test_only:
         for name, dataloader in datamanager.test_loader.items():
-            if not len(dataloader['query'].dataset.classes): # current text annotation doesn't contain classes names
+            if not dataloader['query'].dataset.classes: # current text annotation doesn't contain classes names
                 print(f'Warning: classes are not defined for validation dataset {name}')
                 continue
-            if not len(get_model_attr(model, 'classification_classes')):
-                print(f'Warning: classes are not provided in the current snapshot. Consistency checks are skipped.')
+            if not get_model_attr(model, 'classification_classes'):
+                print('Warning: classes are not provided in the current snapshot. Consistency checks are skipped.')
                 continue
             if not check_classes_consistency(get_model_attr(model, 'classification_classes'),
                                                 dataloader['query'].dataset.classes, strict=False):
                 raise ValueError('Inconsistent classes in evaluation dataset')
             if classes and not check_classes_consistency(classes_map,
-                                                                get_model_attr(model, 'classification_classes'), strict=True):
+                                                         get_model_attr(model, 'classification_classes'), strict=True):
                 raise ValueError('Classes provided via --classes should be the same as in the loaded model')
     elif classes:
         if not check_classes_consistency(classes_map,

torchreid/apis/export.py

@@ -70,7 +70,7 @@ def export_onnx(model, cfg, output_file_path='model', disable_dyn_axes=True,
             onnx.checker.check_model(net_from_onnx)
             print('ONNX check passed.')
         except onnx.onnx_cpp2py_export.checker.ValidationError as ex:
-            print('ONNX check failed: {}.'.format(ex))
+            print(f'ONNX check failed: {ex}.')
 
     return output_file_path
 

torchreid/apis/training.py

@@ -14,6 +14,7 @@
  limitations under the License.
 """
 
+import sys
 from copy import deepcopy
 
 import torchreid
@@ -42,7 +43,7 @@ def run_lr_finder(cfg, datamanager, model, optimizer, scheduler, classes,
     print(f"Estimated learning rate: {aux_lr}")
     if cfg.lr_finder.stop_after:
         print("Finding learning rate finished. Terminate the training process")
-        exit()
+        sys.exit(0)
 
     # reload all parts of the training
     # we do not check classification parameters
@@ -76,7 +77,7 @@ def run_training(cfg, datamanager, model, optimizer, scheduler, extra_device_ids
     num_train_classes = datamanager.num_train_pids
 
     if num_aux_models > 0:
-        print('Enabled mutual learning between {} models.'.format(len(cfg.mutual_learning.aux_configs) + 1))
+        print(f'Enabled mutual learning between {len(cfg.mutual_learning.aux_configs) + 1} models.')
 
         models, optimizers, schedulers = [model], [optimizer], [scheduler]
         for config_file, device_ids in zip(cfg.mutual_learning.aux_configs, extra_device_ids):
@@ -91,7 +92,7 @@ def run_training(cfg, datamanager, model, optimizer, scheduler, extra_device_ids
             schedulers.append(aux_scheduler)
     else:
         models, optimizers, schedulers = model, optimizer, scheduler
-    print('Building {}-engine'.format(cfg.loss.name))
+    print(f'Building {cfg.loss.name}-engine')
     engine = build_engine(cfg, datamanager, models, optimizers, schedulers,
                           should_freeze_aux_models=should_freeze_aux_models,
                           nncf_metainfo=nncf_metainfo,

torchreid/integration/nncf/compression.py

@@ -7,6 +7,7 @@
 
 from torchreid.utils.tools import random_image
 
+# pylint: disable=line-too-long
 
 @contextmanager
 def nullcontext():
@@ -67,7 +68,7 @@ def _load_checkpoint_for_nncf(model, filename, map_location=None, strict=False):
         state_dict = checkpoint['state_dict']
     else:
         raise RuntimeError(
-            'No state_dict found in checkpoint file {}'.format(filename))
+            f'No state_dict found in checkpoint file {filename}')
     _ = load_state(model, state_dict, strict)
     return checkpoint
 

torchreid/integration/nncf/compression_script_utils.py

@@ -6,7 +6,7 @@
 from scripts.default_config import merge_from_files_with_base
 from scripts.script_utils import build_datamanager, is_config_parameter_set_from_command_line
 
-from .compression import (is_checkpoint_nncf, wrap_nncf_model)
+from torchreid.integration.nncf.compression import (is_checkpoint_nncf, wrap_nncf_model)
 
 def get_coeff_decrease_lr_for_nncf(nncf_training_config):
     if nncf_training_config and nncf_training_config.get('coeff_decrease_lr_for_nncf'):
@@ -158,7 +158,7 @@ def make_nncf_changes_in_eval(model, cfg):
     print(f'NNCF config path = {nncf_config_path}')
     checkpoint_path = cfg.model.load_weights
     datamanager_for_nncf = None
-    compression_ctrl, model, _ = \
+    _, model, _ = \
             wrap_nncf_model(model, cfg, datamanager_for_nncf,
                             checkpoint_path=checkpoint_path,
                             nncf_config_path=nncf_config_path)