Update benchmarks.md (PR about issue #3528)

Summary:
PR about issue #3528

Pull Request resolved: #3529

Differential Revision: D31200753

Pulled By: ppwwyyxx

fbshipit-source-id: 84954c59c72ebb0c493c4aa6cbff60befd13e6b2

Author: puhuk

detectron2/config/defaults.py

@@ -557,7 +557,7 @@
 # changing these and they exist only to reproduce Detectron v1 training if
 # desired.
 _C.SOLVER.BIAS_LR_FACTOR = 1.0
-_C.SOLVER.WEIGHT_DECAY_BIAS = None   # None means following WEIGHT_DECAY
+_C.SOLVER.WEIGHT_DECAY_BIAS = None  # None means following WEIGHT_DECAY
 
 # Gradient clipping
 _C.SOLVER.CLIP_GRADIENTS = CN({"ENABLED": False})

detectron2/data/samplers/distributed_sampler.py

@@ -268,7 +268,7 @@ def _get_local_indices(total_size, world_size, rank):
         shard_sizes = [shard_size + int(r < left) for r in range(world_size)]
 
         begin = sum(shard_sizes[:rank])
-        end = min(sum(shard_sizes[:rank + 1]), total_size)
+        end = min(sum(shard_sizes[: rank + 1]), total_size)
         return range(begin, end)
 
     def __iter__(self):

detectron2/utils/visualizer.py

@@ -86,7 +86,10 @@ def __init__(self, mask_or_polygons, height, width):
 
         if isinstance(m, np.ndarray):  # assumed to be a binary mask
             assert m.shape[1] != 2, m.shape
-            assert m.shape == (height, width), f"mask shape: {m.shape}, target dims: {height}, {width}"
+            assert m.shape == (
+                height,
+                width,
+            ), f"mask shape: {m.shape}, target dims: {height}, {width}"
             self._mask = m.astype("uint8")
             return
 
@@ -609,7 +612,7 @@ def overlay_instances(
         masks=None,
         keypoints=None,
         assigned_colors=None,
-        alpha=0.5
+        alpha=0.5,
     ):
         """
         Args:
@@ -852,7 +855,7 @@ def draw_text(
         font_size=None,
         color="g",
         horizontal_alignment="center",
-        rotation=0
+        rotation=0,
     ):
         """
         Args:

docs/notes/benchmarks.md

@@ -12,7 +12,7 @@ with some other popular open source Mask R-CNN implementations.
   TensorFlow 1.15.0rc2, Keras 2.2.5, MxNet 1.6.0b20190820.
 * Model: an end-to-end R-50-FPN Mask-RCNN model, using the same hyperparameter as the
   [Detectron baseline config](https://github.com/facebookresearch/Detectron/blob/master/configs/12_2017_baselines/e2e_mask_rcnn_R-50-FPN_1x.yaml)
-	(it does no have scale augmentation).
+  (it does not have scale augmentation).
 * Metrics: We use the average throughput in iterations 100-500 to skip GPU warmup time.
   Note that for R-CNN-style models, the throughput of a model typically changes during training, because
   it depends on the predictions of the model. Therefore this metric is not directly comparable with
@@ -82,7 +82,7 @@ Details for each implementation:
   ```
 
 * __maskrcnn-benchmark__: use commit `0ce8f6f` with `sed -i 's/torch.uint8/torch.bool/g' **/*.py; sed -i 's/AT_CHECK/TORCH_CHECK/g' **/*.cu`
-	to make it compatible with PyTorch 1.5. Then, run training with
+  to make it compatible with PyTorch 1.5. Then, run training with
   ```
   python -m torch.distributed.launch --nproc_per_node=8 tools/train_net.py --config-file configs/e2e_mask_rcnn_R_50_FPN_1x.yaml
   ```

tests/data/test_sampler.py

@@ -11,9 +11,9 @@
 from detectron2.data.common import DatasetFromList, ToIterableDataset
 from detectron2.data.samplers import (
     GroupedBatchSampler,
+    InferenceSampler,
     RepeatFactorTrainingSampler,
     TrainingSampler,
-    InferenceSampler,
 )
 from detectron2.utils.env import seed_all_rng
 
@@ -102,9 +102,7 @@ def test_local_indices(self):
             [range(11), range(11, 22), range(22, 32), range(32, 42)],
         ]
 
-        for size, world_size, expected_result in zip(
-            sizes, world_sizes, expected_results
-        ):
+        for size, world_size, expected_result in zip(sizes, world_sizes, expected_results):
             with self.subTest(f"size={size}, world_size={world_size}"):
                 local_indices = [
                     InferenceSampler._get_local_indices(size, world_size, r)