Enable back inference mode support with hpu & update links

docs/source-pytorch/accelerators/hpu_basic.rst

@@ -113,4 +113,3 @@ Known limitations
 -----------------
 
 * `Habana dataloader <https://docs.habana.ai/en/latest/PyTorch_User_Guide/PyTorch_User_Guide.html#habana-data-loader>`__ is not supported.
-* :func:`torch.inference_mode` is not supported

src/pytorch_lightning/CHANGELOG.md

@@ -85,6 +85,7 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 
 - Enhanced `reduce_boolean_decision` to accommodate `any`-analogous semantics expected by the `EarlyStopping` callback ([#15253](https://github.com/Lightning-AI/lightning/pull/15253))
 
+- Fixed issue with unsupported torch.inference_mode() on hpu backends ([#15918](https://github.com/Lightning-AI/lightning/pull/15918))
 
 
 ## [1.8.3] - 2022-11-22

src/pytorch_lightning/trainer/connectors/accelerator_connector.py

@@ -80,12 +80,7 @@
 from pytorch_lightning.strategies.ddp_spawn import _DDP_FORK_ALIASES
 from pytorch_lightning.tuner.auto_gpu_select import pick_multiple_gpus
 from pytorch_lightning.utilities.exceptions import MisconfigurationException
-from pytorch_lightning.utilities.imports import (
-    _HOROVOD_AVAILABLE,
-    _HPU_AVAILABLE,
-    _IPU_AVAILABLE,
-    _TORCH_GREATER_EQUAL_1_11,
-)
+from pytorch_lightning.utilities.imports import _HOROVOD_AVAILABLE, _IPU_AVAILABLE, _TORCH_GREATER_EQUAL_1_11
 from pytorch_lightning.utilities.rank_zero import rank_zero_deprecation, rank_zero_info, rank_zero_warn
 
 log = logging.getLogger(__name__)
@@ -499,7 +494,7 @@ def _choose_auto_accelerator(self) -> str:
                 return "tpu"
             if _IPU_AVAILABLE:
                 return "ipu"
-            if _HPU_AVAILABLE:
+            if HPUAccelerator.is_available():
                 return "hpu"
             if MPSAccelerator.is_available():
                 return "mps"

src/pytorch_lightning/trainer/trainer.py

@@ -48,7 +48,7 @@
 from lightning_lite.utilities.data import _auto_add_worker_init_fn
 from lightning_lite.utilities.types import _PATH
 from lightning_lite.utilities.warnings import PossibleUserWarning
-from pytorch_lightning.accelerators import Accelerator, HPUAccelerator, TPUAccelerator
+from pytorch_lightning.accelerators import Accelerator, TPUAccelerator
 from pytorch_lightning.callbacks import Callback, Checkpoint, EarlyStopping, ProgressBarBase
 from pytorch_lightning.callbacks.prediction_writer import BasePredictionWriter
 from pytorch_lightning.core.datamodule import LightningDataModule
@@ -2262,12 +2262,11 @@ def configure_optimizers(self):
 @contextmanager
 def _evaluation_context(accelerator: Accelerator, inference_mode: bool = True) -> Generator:
     # inference mode is not supported with gloo backend (#9431),
-    # and HPU & TPU accelerators.
+    # and TPU accelerators.
     context_manager_class = (
         torch.inference_mode
         if inference_mode
         and not (dist.is_available() and dist.is_initialized() and dist.get_backend() == "gloo")
-        and not isinstance(accelerator, HPUAccelerator)
         and not isinstance(accelerator, TPUAccelerator)
         else torch.no_grad
     )