[Feat] Support training on MPS (#331)

* [Feat] Support mps

* fix docstring

mmengine/device/__init__.py

 # Copyright (c) OpenMMLab. All rights reserved.
 from .utils import (get_device, get_max_cuda_memory, is_cuda_available,
-                    is_mlu_available)
+                    is_mlu_available, is_mps_available)
 
 __all__ = [
     'get_max_cuda_memory', 'get_device', 'is_cuda_available',
-    'is_mlu_available'
+    'is_mlu_available', 'is_mps_available'
 ]

mmengine/device/utils.py

@@ -37,15 +37,25 @@ def is_mlu_available() -> bool:
     return hasattr(torch, 'is_mlu_available') and torch.is_mlu_available()
 
 
+def is_mps_available() -> bool:
+    """Return True if mps devices exist.
+
+    It's specialized for mac m1 chips and require torch version 1.12 or higher.
+    """
+    return hasattr(torch.backends, 'mps') and torch.backends.mps.is_available()
+
+
 def get_device() -> str:
     """Returns the currently existing device type.
 
     Returns:
-        str: cuda | mlu | cpu.
+        str: cuda | mlu | mps | cpu.
     """
     if is_cuda_available():
         return 'cuda'
     elif is_mlu_available():
         return 'mlu'
+    elif is_mps_available():
+        return 'mps'
     else:
         return 'cpu'

tests/test_device/test_device.py

 # Copyright (c) OpenMMLab. All rights reserved.
-from mmengine.device import get_device, is_cuda_available, is_mlu_available
+from mmengine.device import (get_device, is_cuda_available, is_mlu_available,
+                             is_mps_available)
 
 
 def test_get_device():
@@ -8,5 +9,7 @@ def test_get_device():
         assert device == 'cuda'
     elif is_mlu_available():
         assert device == 'mlu'
+    elif is_mps_available():
+        assert device == 'mps'
     else:
         assert device == 'cpu'