Remove Accelerator.parallel_device_ids and deprecate Trainer.data_parallel_device_ids

CHANGELOG.md

@@ -547,6 +547,9 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 - Deprecated `Trainer.num_processes` in favor of `Trainer.num_devices` ([#12388](https://github.com/PyTorchLightning/pytorch-lightning/pull/12388))
 
 
+- Deprecated `Trainer.data_parallel_device_ids` in favor of `Trainer.device_ids` ([#12072](https://github.com/PyTorchLightning/pytorch-lightning/pull/12072))
+
+
 ### Removed
 
 - Removed deprecated parameter `method` in `pytorch_lightning.utilities.model_helpers.is_overridden` ([#10507](https://github.com/PyTorchLightning/pytorch-lightning/pull/10507))
@@ -749,6 +752,9 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 - Removed `AcceleratorConnector.num_processes` property ([#12388](https://github.com/PyTorchLightning/pytorch-lightning/pull/12388))
 
 
+- Removed `AcceleratorConnector.parallel_device_ids` property ([#12072](https://github.com/PyTorchLightning/pytorch-lightning/pull/12072))
+
+
 ### Fixed
 
 - Fixed an issue where `ModelCheckpoint` could delete older checkpoints when `dirpath` has changed during resumed training ([#12045](https://github.com/PyTorchLightning/pytorch-lightning/pull/12045))

pytorch_lightning/callbacks/gpu_stats_monitor.py

@@ -133,8 +133,7 @@ def setup(self, trainer: "pl.Trainer", pl_module: "pl.LightningModule", stage: O
             )
 
         # The logical device IDs for selected devices
-        # ignoring mypy check because `trainer.data_parallel_device_ids` is None when using CPU
-        self._device_ids = sorted(set(trainer.data_parallel_device_ids))  # type: ignore
+        self._device_ids = sorted(set(trainer.device_ids))
 
         # The unmasked real GPU IDs
         self._gpu_ids = self._get_gpu_ids(self._device_ids)

pytorch_lightning/trainer/connectors/accelerator_connector.py

@@ -802,10 +802,6 @@ def tpu_cores(self) -> Optional[Union[List[int], int]]:
     def gpus(self) -> Optional[Union[List[int], str, int]]:
         return self._gpus
 
-    @property
-    def parallel_device_ids(self) -> List[int]:
-        return [i for i in range(len(self.parallel_devices))] if isinstance(self.accelerator, GPUAccelerator) else []
-
     @property
     def is_distributed(self) -> bool:
         # Used for custom plugins.

pytorch_lightning/trainer/connectors/logger_connector/logger_connector.py

@@ -221,7 +221,7 @@ def _log_gpus_metrics(self) -> None:
                 self.trainer.lightning_module.log(key, mem, prog_bar=False, logger=True)
             else:
                 gpu_id = int(key.split("/")[0].split(":")[1])
-                if gpu_id in self.trainer._accelerator_connector.parallel_device_ids:
+                if gpu_id in self.trainer.device_ids:
                     self.trainer.lightning_module.log(
                         key, mem, prog_bar=False, logger=True, on_step=True, on_epoch=False
                     )

pytorch_lightning/trainer/trainer.py

@@ -2114,9 +2114,11 @@ def devices(self) -> int:
 
     @property
     def data_parallel_device_ids(self) -> Optional[List[int]]:
-        return (
-            self._accelerator_connector.parallel_device_ids if self._accelerator_connector.parallel_device_ids else None
+        rank_zero_deprecation(
+            "`Trainer.data_parallel_device_ids` was deprecated in v1.6 and will be removed in v1.8."
+            " Please use `Trainer.device_ids` instead."
         )
+        return self.device_ids if isinstance(self.accelerator, GPUAccelerator) else None
 
     @property
     def lightning_module(self) -> "pl.LightningModule":

tests/deprecated_api/test_remove_1-8.py

@@ -1027,3 +1027,30 @@ def test_trainer_num_processes(monkeypatch, trainer_kwargs, expected_num_process
         "Please use `Trainer.num_devices` instead."
     ):
         trainer.num_processes == expected_num_processes
+
+
+@pytest.mark.parametrize(
+    ["trainer_kwargs", "expected_data_parallel_device_ids"],
+    [
+        ({}, None),
+        ({"devices": 1}, None),
+        ({"devices": "1"}, None),
+        ({"accelerator": "gpu", "devices": 1}, [0]),
+        ({"accelerator": "gpu", "devices": 2}, [0, 1]),
+        ({"accelerator": "gpu", "devices": [1]}, [1]),
+        ({"accelerator": "gpu", "devices": "0"}, None),
+        ({"accelerator": "gpu", "devices": "0,"}, [0]),
+    ],
+)
+def test_trainer_data_parallel_device_ids(monkeypatch, trainer_kwargs, expected_data_parallel_device_ids):
+    """Test multi type argument with bool."""
+    if trainer_kwargs.get("accelerator") == "gpu":
+        monkeypatch.setattr(torch.cuda, "is_available", lambda: True)
+        monkeypatch.setattr(torch.cuda, "device_count", lambda: 2)
+
+    trainer = Trainer(**trainer_kwargs)
+    with pytest.deprecated_call(
+        match="`Trainer.data_parallel_device_ids` was deprecated in v1.6 and will be removed in v1.8."
+        " Please use `Trainer.device_ids` instead."
+    ):
+        assert trainer.data_parallel_device_ids == expected_data_parallel_device_ids

tests/models/test_gpu.py

@@ -184,14 +184,16 @@ def test_parse_gpu_returns_none_when_no_devices_are_available(mocked_device_coun
 )
 @mock.patch("torch.cuda.device_count", return_value=1)
 @mock.patch("torch.cuda.is_available", return_value=True)
-@pytest.mark.parametrize("gpus", [[0, 1, 2], 2, "0"])
+@pytest.mark.parametrize("gpus", [[0, 1, 2], 2, "0", [0, 2]])
 def test_torchelastic_gpu_parsing(mocked_device_count, mocked_is_available, gpus):
     """Ensure when using torchelastic and nproc_per_node is set to the default of 1 per GPU device That we omit
     sanitizing the gpus as only one of the GPUs is visible."""
     trainer = Trainer(gpus=gpus)
     assert isinstance(trainer._accelerator_connector.cluster_environment, TorchElasticEnvironment)
-    assert trainer.data_parallel_device_ids == device_parser.parse_gpu_ids(gpus)
     assert trainer.gpus == gpus
+    # when use gpu
+    if device_parser.parse_gpu_ids(gpus) is not None:
+        assert trainer.device_ids == device_parser.parse_gpu_ids(gpus)
 
 
 @RunIf(min_gpus=1)

tests/trainer/test_trainer_cli.py

@@ -161,10 +161,10 @@ def test_argparse_args_parsing_fast_dev_run(cli_args, expected):
 
 
 @pytest.mark.parametrize(
-    ["cli_args", "expected_parsed", "expected_device_ids"],
-    [("", None, None), ("--accelerator gpu --devices 1", "1", [0]), ("--accelerator gpu --devices 0,", "0,", [0])],
+    ["cli_args", "expected_parsed"],
+    [("", None), ("--accelerator gpu --devices 1", "1"), ("--accelerator gpu --devices 0,", "0,")],
 )
-def test_argparse_args_parsing_devices(cli_args, expected_parsed, expected_device_ids, monkeypatch):
+def test_argparse_args_parsing_devices(cli_args, expected_parsed, monkeypatch):
     """Test multi type argument with bool."""
 
     monkeypatch.setattr(torch.cuda, "is_available", lambda: True)
@@ -177,8 +177,7 @@ def test_argparse_args_parsing_devices(cli_args, expected_parsed, expected_devic
         args = Trainer.parse_argparser(parser)
 
     assert args.devices == expected_parsed
-    trainer = Trainer.from_argparse_args(args)
-    assert trainer.data_parallel_device_ids == expected_device_ids
+    assert Trainer.from_argparse_args(args)
 
 
 @pytest.mark.parametrize(

tests/utilities/test_cli.py

@@ -180,7 +180,9 @@ def test_parse_args_parsing_complex_types(cli_args, expected, instantiate):
         assert Trainer.from_argparse_args(args)
 
 
-@pytest.mark.parametrize(["cli_args", "expected_gpu"], [("--gpus 1", [0]), ("--gpus 0,", [0]), ("--gpus 0,1", [0, 1])])
+@pytest.mark.parametrize(
+    ["cli_args", "expected_gpu"], [("--gpus 1", [0]), ("--gpus 0,", [0]), ("--gpus 1,", [1]), ("--gpus 0,1", [0, 1])]
+)
 def test_parse_args_parsing_gpus(monkeypatch, cli_args, expected_gpu):
     """Test parsing of gpus and instantiation of Trainer."""
     monkeypatch.setattr("torch.cuda.device_count", lambda: 2)
@@ -192,7 +194,7 @@ def test_parse_args_parsing_gpus(monkeypatch, cli_args, expected_gpu):
         args = parser.parse_args()
 
     trainer = Trainer.from_argparse_args(args)
-    assert trainer.data_parallel_device_ids == expected_gpu
+    assert trainer.device_ids == expected_gpu
 
 
 @pytest.mark.skipif(