Customize runner type

configs/_base_/default_runtime.py

@@ -14,3 +14,4 @@
 load_from = None
 resume_from = None
 workflow = [('train', 1)]
+runner = 'EpochBasedRunner'

configs/faster_rcnn/faster_rcnn_r50_caffe_fpn_mstrain_90000it_coco.py

@@ -0,0 +1,16 @@
+_base_ = 'faster_rcnn_r50_caffe_fpn_mstrain_1x_coco.py'
+
+# learning policy
+lr_config = dict(
+ policy='step',
+ warmup='linear',
+ warmup_iters=500,
+ warmup_ratio=0.001,
+ step=[60000, 80000])
+total_epochs = 90000
+
+# Runner type
+runner = 'IterBasedRunner'
+
+checkpoint_config = dict(interval=10000)
+evaluation = dict(interval=10000, metric='bbox')

configs/guided_anchoring/ga_retinanet_r101_caffe_fpn_mstrain_2x.py

@@ -1,3 +1,5 @@
+_base_ = '../_base_/default_runtime.py'
+
 # model settings
 model = dict(
  type='RetinaNet',
@@ -163,9 +165,3 @@
 # yapf:enable
 # runtime settings
 total_epochs = 24
-dist_params = dict(backend='nccl')
-log_level = 'INFO'
-work_dir = './work_dirs/ga_retinanet_r101_caffe_fpn_mstrain_2x'
-load_from = None
-resume_from = None
-workflow = [('train', 1)]

configs/vfnet/vfnet_r50_fpn_1x_coco.py

@@ -106,8 +106,3 @@
  warmup_ratio=0.1,
  step=[8, 11])
 total_epochs = 12
-
-# runtime
-load_from = None
-resume_from = None
-workflow = [('train', 1)]

docs/tutorials/config.md

@@ -409,6 +409,7 @@ load_from = None # load models as a pre-trained model from a given path. This w
 resume_from = None # Resume checkpoints from a given path, the training will be resumed from the epoch when the checkpoint's is saved.
 workflow = [('train', 1)] # Workflow for runner. [('train', 1)] means there is only one workflow and the workflow named 'train' is executed once. The workflow trains the model by 12 epochs according to the total_epochs.
 work_dir = 'work_dir' # Directory to save the model checkpoints and logs for the current experiments.
+runner = 'EpochBasedRunner' # Type of the runner that runs the workflow, can be either 'EpochBasedRunner' or 'IterBasedRunner'.
 
 ```
 

mmdet/apis/train.py

@@ -84,7 +84,9 @@ def train_detector(model,
 
  # build runner
  optimizer = build_optimizer(model, cfg.optimizer)
- runner = EpochBasedRunner(
+ assert cfg.runner in ('IterBasedRunner', 'EpochBasedRunner'), \
+ 'Currently only support IterBasedRunner and EpochBasedRunner'
+ runner = eval(cfg.runner)(
  model,
  optimizer=optimizer,
  work_dir=cfg.work_dir,
@@ -108,7 +110,8 @@ def train_detector(model,
  cfg.checkpoint_config, cfg.log_config,
  cfg.get('momentum_config', None))
  if distributed:
- runner.register_hook(DistSamplerSeedHook())
+ if isinstance(runner, EpochBasedRunner):
+ runner.register_hook(DistSamplerSeedHook())
 
  # register eval hooks
  if validate: