flake8 fixes (#3064)

* flake8 fixes * fix pep8 * fix pep8 Co-authored-by: William Falcon <waf2107@columbia.edu>
Lightning-AI · Aug 20, 2020 · cee5eaf · cee5eaf
1 parent 9a60564
commit cee5eaf
Show file tree

Hide file tree

Showing 5 changed files with 241 additions and 227 deletions.
diff --git a/pytorch_lightning/callbacks/gpu_usage_logger.py b/pytorch_lightning/callbacks/gpu_usage_logger.py
@@ -80,9 +80,15 @@ class GpuUsageLogger(Callback):
 
         """
 
-    def __init__(self, memory_utilisation: bool = True, gpu_utilisation: bool = True,
-                 intra_step_time: bool = False, inter_step_time: bool = False,
-                 fan_speed: bool = False, temperature: bool = False):
+    def __init__(
+        self,
+        memory_utilisation: bool = True,
+        gpu_utilisation: bool = True,
+        intra_step_time: bool = False,
+        inter_step_time: bool = False,
+        fan_speed: bool = False,
+        temperature: bool = False,
+    ):
         super(GpuUsageLogger).__init__()
         self.memory_utilisation = memory_utilisation
         self.gpu_utilisation = gpu_utilisation
@@ -102,9 +108,10 @@ def on_batch_start(self, trainer, pl_module):
         if self.inter_step_time:
             # First log at beginning of second step
             if self.snap_inter_step_time:
-                trainer.logger.log_metrics({'Batch_Time/inter_step (ms)':
-                                                (time.time() - self.snap_inter_step_time) * 1000},
-                                           step=trainer.global_step)
+                trainer.logger.log_metrics(
+                    {'Batch_Time/inter_step (ms)': (time.time() - self.snap_inter_step_time) * 1000},
+                    step=trainer.global_step,
+                )
         if self.intra_step_time:
             self.snap_intra_step_time = time.time()
 
@@ -125,20 +132,24 @@ def on_batch_end(self, trainer, pl_module):
 
         if self.intra_step_time:
             if self.snap_intra_step_time:
-                trainer.logger.log_metrics({'Batch_Time/intra_step (ms)':
-                                                (time.time() - self.snap_intra_step_time) * 1000},
-                                           step=trainer.global_step)
+                trainer.logger.log_metrics(
+                    {'Batch_Time/intra_step (ms)': (time.time() - self.snap_intra_step_time) * 1000},
+                    step=trainer.global_step,
+                )
 
     def on_train_epoch_start(self, trainer, pl_module):
         self.snap_intra_step_time = None
         self.snap_inter_step_time = None
 
     @staticmethod
     def _get_gpu_stat(pitem: str, unit: str):
-        result = subprocess.run(["/bin/nvidia-smi", f"--query-gpu={pitem}", "--format=csv,nounits,noheader"],
-                                encoding="utf-8", stdout=subprocess.PIPE,
-                                stderr=subprocess.PIPE,  # for backward compatibility with python version 3.6
-                                check=True)
+        result = subprocess.run(
+            ["/bin/nvidia-smi", f"--query-gpu={pitem}", "--format=csv,nounits,noheader"],
+            encoding="utf-8",
+            stdout=subprocess.PIPE,
+            stderr=subprocess.PIPE,  # for backward compatibility with python version 3.6
+            check=True,
+        )
         try:
             gpu_usage = [float(x) for x in result.stdout.strip().split(os.linesep)]
         except ValueError:
@@ -152,4 +163,4 @@ def _log_gpu(self, trainer):
     def _log_memory(self, trainer):
         trainer.logger.log_metrics(self._get_gpu_stat("memory.used", "MB"), step=trainer.global_step)
         trainer.logger.log_metrics(self._get_gpu_stat("memory.free", "MB"), step=trainer.global_step)
-        trainer.logger.log_metrics(self._get_gpu_stat("utilization.memory", "%"), step=trainer.global_step)
+        trainer.logger.log_metrics(self._get_gpu_stat("utilization.memory", "%"), step=trainer.global_step)
diff --git a/pytorch_lightning/core/memory.py b/pytorch_lightning/core/memory.py
@@ -222,7 +222,7 @@ def _forward_example_input(self) -> None:
         input_ = model.transfer_batch_to_device(input_, model.device)
 
         if trainer is not None and trainer.amp_backend == AMPType.NATIVE and not trainer.use_tpu:
-                model.forward = torch.cuda.amp.autocast()(model.forward)
+            model.forward = torch.cuda.amp.autocast()(model.forward)
 
         mode = model.training
         model.eval()