[Enhancement] Support deepspeed with flexible runner (open-mmlab#1673)

* [Feature] Support deepspeed with flexible runner

* [Fix] Reformat with yapf

* [Refacor] Rename configs

* [Fix] Reformat with yapf

* [Refactor] Remove unused keys

* [Refactor] Change the _base_ path

* [Refactor] Reformat

configs/mae/benchmarks/vit-huge-p14_8xb128-ds-zero3-coslr-50e_in1k.py

@@ -0,0 +1,32 @@
+_base_ = ['./vit-huge-p14_8xb128-coslr-50e_in1k.py']
+
+# optimizer wrapper
+optim_wrapper = dict(type='DeepSpeedOptimWrapper')
+
+# training strategy
+# Deepspeed with ZeRO3 + fp16
+strategy = dict(
+    type='DeepSpeedStrategy',
+    fp16=dict(
+        enabled=True,
+        fp16_master_weights_and_grads=False,
+        loss_scale=0,
+        loss_scale_window=500,
+        hysteresis=2,
+        min_loss_scale=1,
+        initial_scale_power=15,
+    ),
+    inputs_to_half=['inputs'],
+    zero_optimization=dict(
+        stage=3,
+        allgather_partitions=True,
+        reduce_scatter=True,
+        allgather_bucket_size=50000000,
+        reduce_bucket_size=50000000,
+        overlap_comm=True,
+        contiguous_gradients=True,
+        cpu_offload=False,
+    ))
+
+# runner which supports strategies
+runner_type = 'FlexibleRunner'

configs/mae/benchmarks/vit-large-p16_8xb128-ds-zero3-coslr-50e_in1k.py

@@ -0,0 +1,32 @@
+_base_ = ['./vit-large-p16_8xb128-coslr-50e_in1k.py']
+
+# optimizer wrapper
+optim_wrapper = dict(type='DeepSpeedOptimWrapper')
+
+# training strategy
+# Deepspeed with ZeRO3 + fp16
+strategy = dict(
+    type='DeepSpeedStrategy',
+    fp16=dict(
+        enabled=True,
+        fp16_master_weights_and_grads=False,
+        loss_scale=0,
+        loss_scale_window=500,
+        hysteresis=2,
+        min_loss_scale=1,
+        initial_scale_power=15,
+    ),
+    inputs_to_half=['inputs'],
+    zero_optimization=dict(
+        stage=3,
+        allgather_partitions=True,
+        reduce_scatter=True,
+        allgather_bucket_size=50000000,
+        reduce_bucket_size=50000000,
+        overlap_comm=True,
+        contiguous_gradients=True,
+        cpu_offload=False,
+    ))
+
+# runner which supports strategies
+runner_type = 'FlexibleRunner'

tools/test.py

@@ -7,6 +7,7 @@
 import mmengine
 from mmengine.config import Config, ConfigDict, DictAction
 from mmengine.evaluator import DumpResults
+from mmengine.registry import RUNNERS
 from mmengine.runner import Runner
 
 
@@ -169,7 +170,13 @@ def main():
     cfg = merge_args(cfg, args)
 
     # build the runner from config
-    runner = Runner.from_cfg(cfg)
+    if 'runner_type' not in cfg:
+        # build the default runner
+        runner = Runner.from_cfg(cfg)
+    else:
+        # build customized runner from the registry
+        # if 'runner_type' is set in the cfg
+        runner = RUNNERS.build(cfg)
 
     if args.out and args.out_item in ['pred', None]:
         runner.test_evaluator.metrics.append(

tools/train.py

@@ -5,6 +5,7 @@
 from copy import deepcopy
 
 from mmengine.config import Config, ConfigDict, DictAction
+from mmengine.registry import RUNNERS
 from mmengine.runner import Runner
 from mmengine.utils import digit_version
 from mmengine.utils.dl_utils import TORCH_VERSION
@@ -149,7 +150,13 @@ def main():
     cfg = merge_args(cfg, args)
 
     # build the runner from config
-    runner = Runner.from_cfg(cfg)
+    if 'runner_type' not in cfg:
+        # build the default runner
+        runner = Runner.from_cfg(cfg)
+    else:
+        # build customized runner from the registry
+        # if 'runner_type' is set in the cfg
+        runner = RUNNERS.build(cfg)
 
     # start training
     runner.train()