Fix deepspeed default precision plugin amp_level to O2

src/pytorch_lightning/CHANGELOG.md

@@ -393,6 +393,9 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 - Fixed an issue that caused the learning rate finder to set the model's learning rate to None when no suggestion was possible ([#13845](https://github.com/Lightning-AI/lightning/pull/13845))
 
 
+- Fixed default `amp_level` for `DeepSpeedPrecisionPlugin` to `O2` ([#13897](https://github.com/PyTorchLightning/pytorch-lightning/pull/13897))
+
+
 
 ## [1.6.5] - 2022-07-13
 

src/pytorch_lightning/plugins/precision/apex_amp.py

@@ -35,7 +35,7 @@ class ApexMixedPrecisionPlugin(MixedPrecisionPlugin):
     def __init__(self, amp_level: str = "O2") -> None:
         if not _APEX_AVAILABLE:
             raise MisconfigurationException(
-                "You have asked for Apex AMP but you have not installed it."
+                "You have asked for Apex AMP but `apex` is not installed."
                 " Install `apex` using this guide: https://github.com/NVIDIA/apex"
             )
         super().__init__()

src/pytorch_lightning/plugins/precision/deepspeed.py

@@ -20,9 +20,9 @@
 import pytorch_lightning as pl
 from pytorch_lightning.plugins.precision.precision_plugin import PrecisionPlugin
 from pytorch_lightning.utilities import GradClipAlgorithmType
-from pytorch_lightning.utilities.enums import PrecisionType
+from pytorch_lightning.utilities.enums import AMPType, PrecisionType
 from pytorch_lightning.utilities.exceptions import MisconfigurationException
-from pytorch_lightning.utilities.imports import _RequirementAvailable
+from pytorch_lightning.utilities.imports import _APEX_AVAILABLE, _RequirementAvailable
 from pytorch_lightning.utilities.model_helpers import is_overridden
 from pytorch_lightning.utilities.warnings import WarningCache
 
@@ -51,6 +51,15 @@ class DeepSpeedPrecisionPlugin(PrecisionPlugin):
     """
 
     def __init__(self, precision: Union[str, int], amp_type: str, amp_level: Optional[str] = None) -> None:
+        if amp_type == AMPType.APEX:
+            if not _APEX_AVAILABLE:
+                raise MisconfigurationException(
+                    "You have asked for Apex AMP but `apex` is not installed."
+                    " Install `apex` using this guide: https://github.com/NVIDIA/apex"
+                )
+
+            amp_level = amp_level or "O2"
+
         supported_precision = (PrecisionType.HALF, PrecisionType.FLOAT, PrecisionType.BFLOAT, PrecisionType.MIXED)
         if precision not in supported_precision:
             raise ValueError(

src/pytorch_lightning/trainer/connectors/accelerator_connector.py

@@ -185,7 +185,7 @@ def __init__(
         self._layer_sync: Optional[LayerSync] = NativeSyncBatchNorm() if sync_batchnorm else None
         self.checkpoint_io: Optional[CheckpointIO] = None
         self._amp_type_flag: Optional[LightningEnum] = None
-        self._amp_level_flag: Optional[str] = amp_level
+        self._amp_level_flag: str = amp_level
         self._auto_select_gpus: bool = auto_select_gpus
 
         self._check_config_and_set_final_flags(

tests/tests_pytorch/plugins/precision/test_deepspeed_precision.py

@@ -11,11 +11,29 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+from unittest import mock
+
 import pytest
 
 from pytorch_lightning.plugins.precision.deepspeed import DeepSpeedPrecisionPlugin
+from pytorch_lightning.utilities.exceptions import MisconfigurationException
 
 
 def test_invalid_precision_with_deepspeed_precision():
     with pytest.raises(ValueError, match="is not supported. `precision` must be one of"):
         DeepSpeedPrecisionPlugin(precision=64, amp_type="native")
+
+
+def test_deepspeed_precision_apex_not_installed(monkeypatch):
+    import pytorch_lightning.plugins.precision.deepspeed as deepspeed_apex
+
+    monkeypatch.setattr(deepspeed_apex, "_APEX_AVAILABLE", False)
+    with pytest.raises(MisconfigurationException, match="You have asked for Apex AMP but `apex` is not installed."):
+        DeepSpeedPrecisionPlugin(precision=16, amp_type="apex")
+
+
+@mock.patch("pytorch_lightning.plugins.precision.deepspeed._APEX_AVAILABLE", return_value=True)
+def test_deepspeed_precision_apex_default_level(_):
+    precision_plugin = DeepSpeedPrecisionPlugin(precision=16, amp_type="apex")
+    assert isinstance(precision_plugin, DeepSpeedPrecisionPlugin)
+    assert precision_plugin.amp_level == "O2"

tests/tests_pytorch/plugins/test_amp_plugins.py

@@ -289,5 +289,5 @@ def test_precision_selection_raises(monkeypatch):
     monkeypatch.setattr(apex, "_APEX_AVAILABLE", False)
     with mock.patch("pytorch_lightning.utilities.device_parser.num_cuda_devices", return_value=1), mock.patch(
         "pytorch_lightning.utilities.device_parser.is_cuda_available", return_value=True
-    ), pytest.raises(MisconfigurationException, match="asked for Apex AMP but you have not installed it"):
+    ), pytest.raises(MisconfigurationException, match="asked for Apex AMP but `apex` is not installed"):
         Trainer(amp_backend="apex", precision=16, accelerator="gpu", devices=1)