Add MPS accelerator support for mixed precision

src/lightning/pytorch/CHANGELOG.md

@@ -30,6 +30,9 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 - Default to `RichProgressBar` and `RichModelSummary` if the rich package is available. Fallback to TQDMProgressBar and ModelSummary otherwise ([#20896](https://github.com/Lightning-AI/pytorch-lightning/pull/20896))
 
 
+- Add MPS accelerator support for mixed precision ([#21209](https://github.com/Lightning-AI/pytorch-lightning/pull/21209))
+
+
 ### Removed
 
 -

src/lightning/pytorch/trainer/connectors/accelerator_connector.py

@@ -515,7 +515,7 @@ def _check_and_init_precision(self) -> Precision:
             rank_zero_info(
                 f"Using {'16bit' if self._precision_flag == '16-mixed' else 'bfloat16'} Automatic Mixed Precision (AMP)"
             )
-            device = "cpu" if self._accelerator_flag == "cpu" else "cuda"
+            device = self._accelerator_flag if self._accelerator_flag in ("cpu", "mps") else "cuda"
             return MixedPrecision(self._precision_flag, device)  # type: ignore[arg-type]
 
         raise RuntimeError("No precision set")

tests/tests_pytorch/trainer/connectors/test_accelerator_connector.py

@@ -1084,3 +1084,19 @@ def test_precision_selection_model_parallel(precision, raises, mps_count_0):
     error_context = pytest.raises(ValueError, match=f"does not support .*{precision}") if raises else nullcontext()
     with error_context:
         _AcceleratorConnector(precision=precision, strategy=ModelParallelStrategy())
+
+
+@RunIf(mps=True)
+@pytest.mark.parametrize(
+    ("accelerator", "expected_device"),
+    [
+        ("mps", "mps"),
+        ("cpu", "cpu"),
+    ],
+)
+@pytest.mark.parametrize("precision", ["16-mixed", "bf16-mixed"])
+def test_mps_amp_device_selection(accelerator, expected_device, precision):
+    """Test that MPS accelerator with mixed precision correctly sets device to 'mps' instead of 'cuda'."""
+    connector = _AcceleratorConnector(accelerator=accelerator, precision=precision)
+    assert isinstance(connector.precision_plugin, MixedPrecision)
+    assert connector.precision_plugin.device == expected_device