Add assertions to IPU accelerator for device availability

CHANGELOG.md

@@ -98,6 +98,9 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 - Added a `_Stateful` support for `LightningDataModule` ([#11637](https://github.com/PyTorchLightning/pytorch-lightning/pull/11637))
 
 
+- Added checks to `IPUAccelerator` to assert device availability ([#11799](https://github.com/PyTorchLightning/pytorch-lightning/pull/11798))
+
+
 ### Changed
 
 - Implemented a new native and rich format in `_print_results` method of the `EvaluationLoop` ([#11332](https://github.com/PyTorchLightning/pytorch-lightning/pull/11332))

pytorch_lightning/accelerators/ipu.py

@@ -16,11 +16,25 @@
 import torch
 
 from pytorch_lightning.accelerators.accelerator import Accelerator
+from pytorch_lightning.utilities.exceptions import MisconfigurationException
+from pytorch_lightning.utilities.imports import _IPU_AVAILABLE
 
 
 class IPUAccelerator(Accelerator):
     """Accelerator for IPUs."""
 
+    def __init__(self) -> None:
+        """
+        Raises:
+            MisconfigurationException:
+                If the IPU device is not available.
+        """
+        if not _IPU_AVAILABLE:
+            raise MisconfigurationException(
+                "The IPU Accelerator requires IPU devices to run. "
+                "Learn more or get started with IPUs at https://www.graphcore.ai/getstarted"
+            )
+
     def get_device_stats(self, device: Union[str, torch.device]) -> Dict[str, Any]:
         """IPU device stats aren't supported yet."""
         return {}