From f03fd9b276bfba1ec87a640c01172e41d816fe07 Mon Sep 17 00:00:00 2001
From: Saurabh Mishra <msaurabh@meta.com>
Date: Tue, 24 Sep 2024 08:19:56 -0700
Subject: [PATCH] Update Checkpoint docs with DCP based checkpointer (#904)

Summary:
Pull Request resolved: https://github.com/pytorch/tnt/pull/904

Update Checkpoint docs with DCP based checkpointer

Reviewed By: JKSenthil

Differential Revision: D63278746

fbshipit-source-id: 0c01e21fb516996001d3c12fe74504f65f5ed783
---
 docs/source/checkpointing.rst | 30 +++++++++++++++---------------
 1 file changed, 15 insertions(+), 15 deletions(-)

diff --git a/docs/source/checkpointing.rst b/docs/source/checkpointing.rst
index 9596c0d776..466b4b262e 100644
--- a/docs/source/checkpointing.rst
+++ b/docs/source/checkpointing.rst
@@ -1,24 +1,24 @@
 Checkpointing
 ================================
 
-TorchTNT offers checkpointing via the :class:`~torchtnt.framework.callbacks.TorchSnapshotSaver` which uses `TorchSnapshot <https://pytorch.org/torchsnapshot/main/>`_ under the hood.
+TorchTNT offers checkpointing via :class:`~torchtnt.framework.callbacks.DistributedCheckpointSaver` which uses `DCP <https://github.com/pytorch/pytorch/tree/main/torch/distributed/checkpoint>`_ under the hood.
 
 .. code-block:: python
 
     module = nn.Linear(input_dim, 1)
     unit = MyUnit(module=module)
-    tss = TorchSnapshotSaver(
+    dcp = DistributedCheckpointSaver(
         dirpath=your_dirpath_here,
         save_every_n_train_steps=100,
         save_every_n_epochs=2,
     )
     # loads latest checkpoint, if it exists
     if latest_checkpoint_dir:
-        tss.restore_from_latest(your_dirpath_here, unit, train_dataloader=dataloader)
+        dcp.restore_from_latest(your_dirpath_here, unit, train_dataloader=dataloader)
     train(
         unit,
         dataloader,
-        callbacks=[tss]
+        callbacks=[dcp]
     )
 
 There is built-in support for saving and loading distributed models (DDP, FSDP).
@@ -37,7 +37,7 @@ The state dict type to be used for checkpointing FSDP modules can be specified i
     )
     module = prepare_fsdp(module, strategy=fsdp_strategy)
     unit = MyUnit(module=module)
-    tss = TorchSnapshotSaver(
+    dcp = DistributedCheckpointSaver(
         dirpath=your_dirpath_here,
         save_every_n_epochs=2,
     )
@@ -45,7 +45,7 @@ The state dict type to be used for checkpointing FSDP modules can be specified i
         unit,
         dataloader,
         # checkpointer callback will use state dict type specified in FSDPStrategy
-        callbacks=[tss]
+        callbacks=[dcp]
     )
 
 Or you can manually set this using `FSDP.set_state_dict_type <https://pytorch.org/docs/stable/fsdp.html#torch.distributed.fsdp.FullyShardedDataParallel.set_state_dict_type>`_.
@@ -56,14 +56,14 @@ Or you can manually set this using `FSDP.set_state_dict_type <https://pytorch.or
     module = FSDP(module, ....)
     FSDP.set_state_dict_type(module, StateDictType.SHARDED_STATE_DICT)
     unit = MyUnit(module=module, ...)
-    tss = TorchSnapshotSaver(
+    dcp = DistributedCheckpointSaver(
         dirpath=your_dirpath_here,
         save_every_n_epochs=2,
     )
     train(
         unit,
         dataloader,
-        callbacks=[tss]
+        callbacks=[dcp]
     )
 
 
@@ -74,7 +74,7 @@ When finetuning your models, you can pass RestoreOptions to avoid loading optimi
 
 .. code-block:: python
 
-    tss = TorchSnapshotSaver(
+    dcp = DistributedCheckpointSaver(
         dirpath=your_dirpath_here,
         save_every_n_train_steps=100,
         save_every_n_epochs=2,
@@ -82,7 +82,7 @@ When finetuning your models, you can pass RestoreOptions to avoid loading optimi
 
     # loads latest checkpoint, if it exists
     if latest_checkpoint_dir:
-        tss.restore_from_latest(
+        dcp.restore_from_latest(
             your_dirpath_here,
             your_unit,
             train_dataloader=dataloader,
@@ -99,7 +99,7 @@ Sometimes it may be helpful to keep track of how models perform. This can be don
 
     module = nn.Linear(input_dim, 1)
     unit = MyUnit(module=module)
-    tss = TorchSnapshotSaver(
+    dcp = DistributedCheckpointSaver(
         dirpath=your_dirpath_here,
         save_every_n_epochs=1,
         best_checkpoint_config=BestCheckpointConfig(
@@ -111,7 +111,7 @@ Sometimes it may be helpful to keep track of how models perform. This can be don
     train(
         unit,
         dataloader,
-        callbacks=[tss]
+        callbacks=[dcp]
     )
 
 By specifying the monitored metric to be "train_loss", the checkpointer will expect the :class:`~torchtnt.framework.unit.TrainUnit` to have a "train_loss" attribute at the time of checkpointing, and it will cast this value to a float and append the value to the checkpoint path name. This attribute is expected to be computed and kept up to date appropriately in the unit by the user.
@@ -120,13 +120,13 @@ Later on, the best checkpoint can be loaded via
 
 .. code-block:: python
 
-    TorchSnapshotSaver.restore_from_best(your_dirpath_here, unit, metric_name="train_loss", mode="min")
+    DistributedCheckpointSaver.restore_from_best(your_dirpath_here, unit, metric_name="train_loss", mode="min")
 
 If you'd like to monitor a validation metric (say validation loss after each eval epoch during :py:func:`~torchtnt.framework.fit.fit`), you can use the `save_every_n_eval_epochs` flag instead, like so
 
 .. code-block:: python
 
-    tss = TorchSnapshotSaver(
+    dcp = DistributedCheckpointSaver(
         dirpath=your_dirpath_here,
         save_every_n_eval_epochs=1,
         best_checkpoint_config=BestCheckpointConfig(
@@ -139,7 +139,7 @@ And to save only the top three performing models, you can use the existing `keep
 
 .. code-block:: python
 
-    tss = TorchSnapshotSaver(
+    dcp = DistributedCheckpointSaver(
         dirpath=your_dirpath_here,
         save_every_n_eval_epochs=1,
         keep_last_n_checkpoints=3,