[DCP] Minor refactor to avoid passing the epoch in the checkpoint sav…

…e APIs

recipes/full_finetune_distributed.py

@@ -701,10 +701,7 @@ def _setup_data(
 
         return sampler, dataloader
 
-    def save_checkpoint_async(
-        self,
-        epoch: int,
-    ) -> None:
+    def save_checkpoint_async(self) -> None:
         """
         Checkpoint the state of the recipe. The constructed checkpoint state dict
         contains the following information:
@@ -746,7 +743,7 @@ def save_checkpoint_async(
 
         dcp_saver.save_checkpoint(
             ckpt_dict,
-            epoch=epoch,
+            epoch=self.epochs_run,
             save_async=True,
         )
 
@@ -755,10 +752,7 @@ def save_checkpoint_async(
                 f"Saving asynchronous checkpoint took {time.perf_counter() - cp_start:.2f} secs"
             )
 
-    def save_checkpoint(
-        self,
-        epoch: int,
-    ) -> None:
+    def save_checkpoint(self) -> None:
         """
         Checkpoint the state of the recipe. The constructed checkpoint state dict
         contains the following information:
@@ -772,12 +766,12 @@ def save_checkpoint(
         # final dict passed onto the checkpointer
         checkpoint_dict = {}
 
-        intermediate_checkpoint = epoch + 1 < self.total_epochs
+        intermediate_checkpoint = self.epochs_run < self.total_epochs
 
         # If async checkpointing is enabled, intermediate checkpoints will
         # be saved asynchronously.
         if intermediate_checkpoint and self._enable_async_checkpointing:
-            self.save_checkpoint_async(epoch)
+            self.save_checkpoint_async()
             return
 
         if self._is_rank_zero:
@@ -845,7 +839,7 @@ def save_checkpoint(
 
             self._checkpointer.save_checkpoint(
                 checkpoint_dict,
-                epoch=epoch,
+                epoch=self.epochs_run,
                 intermediate_checkpoint=intermediate_checkpoint,
             )
 
@@ -1023,7 +1017,8 @@ def train(self) -> None:
                     self._profiler.step()
 
             self.epochs_run += 1
-            self.save_checkpoint(epoch=curr_epoch)
+            # Save the checkpoint for the current epoch
+            self.save_checkpoint()
 
         self._profiler.stop()