fix make doc and update precision reference

Author: four4fish

pytorch_lightning/accelerators/accelerator.py

@@ -280,10 +280,9 @@ def amp_backend(self) -> Optional[LightningEnum]:
 
     @property
     def precision(self) -> Union[str, int]:
-        """
-        .. deprecated
-            This method is deprecated will be removed soon.
-            Use :`training_type_plugin.precision_plugin.precision` instead.
+        """This method is deprecated and will be removed soon.
+
+        Use `training_type_plugin.precision_plugin.precision` instead.
         """
         rank_zero_deprecation(
             f"`{self.__class__.__name__}.precision` was and will be removed soon"

pytorch_lightning/plugins/training_type/ipu.py

@@ -118,7 +118,7 @@ def setup(self) -> None:
         self.lightning_module.trainer._update_dataloader = self._convert_to_poptorch_loader
 
     def pre_dispatch(self) -> None:
-        model = LightningIPUModule(self.lightning_module, self.precision)
+        model = LightningIPUModule(self.lightning_module, self.precision_plugin.precision)
         self.model = model
 
         # reset the backup

pytorch_lightning/plugins/training_type/sharded.py

@@ -75,7 +75,7 @@ def _reinit_optimizers_with_oss(self, optimizers: List[Union[Optimizer, Lightnin
                 optim_class = type(optimizer)
                 zero_optimizer = OSS(params=optimizer.param_groups, optim=optim_class, **optimizer.defaults)
                 if _FAIRSCALE_OSS_FP16_BROADCAST_AVAILABLE:
-                    precision = self._precision or self.precision
+                    precision = self._precision or self.precision_plugin.precision
                     is_fp16 = precision in ("mixed", 16)
                     # For multi-node training, compressing the model shards in fp16 before broadcasting
                     # improves performance. When using PyTorch AMP, it will not degrade