fix config dependence & fix python 3.8/3.9 install (#2987)

* fix config dependence & fix python 3.8/3.9 install

ppdet/engine/callbacks.py

@@ -217,7 +217,7 @@ class VisualDLWriter(Callback):
             logger.error('visualdl not found, plaese install visualdl. '
                          'for example: `pip install visualdl`.')
             raise e
-        self.vdl_writer = LogWriter(model.cfg.vdl_log_dir)
+        self.vdl_writer = LogWriter(model.cfg.get('vdl_log_dir', 'vdl_log_dir/scalar'))
         self.vdl_loss_step = 0
         self.vdl_mAP_step = 0
         self.vdl_image_step = 0

ppdet/engine/trainer.py

@@ -116,7 +116,7 @@ class Trainer(object):
     def _init_callbacks(self):
         if self.mode == 'train':
             self._callbacks = [LogPrinter(self), Checkpointer(self)]
-            if 'use_vdl' in self.cfg and self.cfg.use_vdl:
+            if self.cfg.get('use_vdl', False):
                 self._callbacks.append(VisualDLWriter(self))
             self._compose_callback = ComposeCallback(self._callbacks)
         elif self.mode == 'eval':
@@ -124,7 +124,7 @@ class Trainer(object):
             if self.cfg.metric == 'WiderFace':
                 self._callbacks.append(WiferFaceEval(self))
             self._compose_callback = ComposeCallback(self._callbacks)
-        elif self.mode == 'test' and 'use_vdl' in self.cfg and self.cfg.use_vdl:
+        elif self.mode == 'test' and self.cfg.get('use_vdl', False):
             self._callbacks = [VisualDLWriter(self)]
             self._compose_callback = ComposeCallback(self._callbacks)
         else:
@@ -141,8 +141,7 @@ class Trainer(object):
             bias = self.cfg['bias'] if 'bias' in self.cfg else 0
             output_eval = self.cfg['output_eval'] \
                 if 'output_eval' in self.cfg else None
-            save_prediction_only = self.cfg['save_prediction_only'] \
-                if 'save_prediction_only' in self.cfg else False
+            save_prediction_only = self.cfg.get('save_prediction_only', False)
 
             # pass clsid2catid info to metric instance to avoid multiple loading
             # annotation file
@@ -253,7 +252,7 @@ class Trainer(object):
             self._reset_metrics()
 
         model = self.model
-        if self.cfg.fleet:
+        if self.cfg.get('fleet', False):
             model = fleet.distributed_model(model)
             self.optimizer = fleet.distributed_optimizer(
                 self.optimizer).user_defined_optimizer
@@ -264,7 +263,7 @@ class Trainer(object):
                 self.model, find_unused_parameters=find_unused_parameters)
 
         # initial fp16
-        if self.cfg.fp16:
+        if self.cfg.get('fp16', False):
             scaler = amp.GradScaler(
                 enable=self.cfg.use_gpu, init_loss_scaling=1024)
 
@@ -292,7 +291,7 @@ class Trainer(object):
                 self.status['step_id'] = step_id
                 self._compose_callback.on_step_begin(self.status)
 
-                if self.cfg.fp16:
+                if self.cfg.get('fp16', False):
                     with amp.auto_cast(enable=self.cfg.use_gpu):
                         # model forward
                         outputs = model(data)

requirements.txt

 tqdm
 typeguard ; python_version >= '3.4'
-visualdl>=2.1.0
+visualdl>=2.1.0 ; python_version <= '3.7'
 opencv-python
 PyYAML
 shapely

setup.py

@@ -87,6 +87,8 @@ if __name__ == "__main__":
             'Programming Language :: Python :: 3.5',
             'Programming Language :: Python :: 3.6',
             'Programming Language :: Python :: 3.7', 'Topic :: Utilities'
+            'Programming Language :: Python :: 3.8', 'Topic :: Utilities'
+            'Programming Language :: Python :: 3.9', 'Topic :: Utilities'
         ],
         license='Apache License 2.0',
         ext_modules=[])