[checkpointio] support async model save (#6131)

* [checkpointio] support async model save

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

Author: ver217

colossalai/booster/booster.py

@@ -310,6 +310,7 @@ def save_model(
         prefix: Optional[str] = None,
         size_per_shard: int = 1024,
         use_safetensors: bool = False,
+        use_async: bool = False,
     ) -> None:
         """Save model to checkpoint.
 
@@ -333,6 +334,7 @@ def save_model(
             prefix=prefix,
             size_per_shard=size_per_shard,
             use_safetensors=use_safetensors,
+            use_async=use_async,
         )
 
     def load_optimizer(self, optimizer: Optimizer, checkpoint: str) -> None:

colossalai/booster/plugin/low_level_zero_plugin.py

@@ -259,10 +259,12 @@ def load_sharded_model(
         super().load_sharded_model(model, checkpoint_index_file, strict, use_safetensors, load_sub_module)
         model.update_master_params()
 
-    def save_unsharded_model(self, model: ModelWrapper, checkpoint: str, gather_dtensor: bool, use_safetensors: bool):
+    def save_unsharded_model(
+        self, model: ModelWrapper, checkpoint: str, gather_dtensor: bool, use_safetensors: bool, use_async: bool = False
+    ):
         assert isinstance(model, LowLevelZeroModel), "Please boost the model before loading!"
         model._force_wait_all_gather()
-        return super().save_unsharded_model(model, checkpoint, gather_dtensor, use_safetensors)
+        return super().save_unsharded_model(model, checkpoint, gather_dtensor, use_safetensors, use_async=use_async)
 
     def save_sharded_model(
         self,
@@ -272,11 +274,12 @@ def save_sharded_model(
         prefix: Optional[str] = None,
         max_shard_size: int = 1024,
         use_safetensors: bool = False,
+        use_async: bool = False,
     ):
         assert isinstance(model, LowLevelZeroModel), "Please boost the model before loading!"
         model._force_wait_all_gather()
         return super().save_sharded_model(
-            model, checkpoint_path, gather_dtensor, prefix, max_shard_size, use_safetensors
+            model, checkpoint_path, gather_dtensor, prefix, max_shard_size, use_safetensors, use_async=use_async
         )
 
     def save_lora_as_pretrained(self, model, checkpoint, use_safetensors):

colossalai/booster/plugin/torch_ddp_plugin.py

@@ -33,13 +33,17 @@ def load_unsharded_model(self, model: ModelWrapper, checkpoint: str, strict: boo
         assert isinstance(model, ModelWrapper), "Please boost the model before loading!"
         super().load_unsharded_model(model.unwrap(), checkpoint, strict=strict)
 
-    def save_unsharded_model(self, model: ModelWrapper, checkpoint: str, gather_dtensor: bool, use_safetensors: bool):
+    def save_unsharded_model(
+        self, model: ModelWrapper, checkpoint: str, gather_dtensor: bool, use_safetensors: bool, use_async: bool = False
+    ):
         """
         Save model to checkpoint but only on master process.
         """
         assert isinstance(model, ModelWrapper), "Please boost the model before saving!"
         if self.coordinator.is_master():
-            super().save_unsharded_model(model.unwrap(), checkpoint, gather_dtensor, use_safetensors)
+            super().save_unsharded_model(
+                model.unwrap(), checkpoint, gather_dtensor, use_safetensors, use_async=use_async
+            )
 
     def load_unsharded_optimizer(self, optimizer: OptimizerWrapper, checkpoint: str):
         """
@@ -71,14 +75,21 @@ def save_sharded_model(
         prefix: Optional[str] = None,
         max_shard_size: int = 1024,
         use_safetensors: bool = False,
+        use_async: bool = False,
     ):
         """
         Save model to checkpoint but only on master process.
         """
         assert isinstance(model, ModelWrapper), "Please boost the model before saving!"
         if self.coordinator.is_master():
             super().save_sharded_model(
-                model.unwrap(), checkpoint_path, gather_dtensor, prefix, max_shard_size, use_safetensors
+                model.unwrap(),
+                checkpoint_path,
+                gather_dtensor,
+                prefix,
+                max_shard_size,
+                use_safetensors,
+                use_async=use_async,
             )
 
     def load_sharded_model(

colossalai/checkpoint_io/checkpoint_io_base.py

@@ -1,13 +1,14 @@
 from abc import ABC, abstractmethod
 from pathlib import Path
-from typing import Optional, Union
+from typing import Dict, Optional, Union
 
 import torch
 import torch.nn as nn
 from torch.optim import Optimizer
 from torch.optim.lr_scheduler import _LRScheduler as LRScheduler
 
 from colossalai.interface import ModelWrapper
+from colossalai.logging import get_dist_logger
 
 from .utils import SAFE_WEIGHTS_NAME, WEIGHTS_NAME, has_index_file
 
@@ -58,9 +59,34 @@ class CheckpointIO(ABC):
         >>> checkpoint_io.save_optimizer(optimizer, 'optimizer.pt')
     """
 
+    N_WRITE_ENTRIES: int = 32
+
     # ======================================
     # Public methods
     # ======================================
+    def __init__(self):
+        super().__init__()
+        self.pinned_state_dicts: Dict[int, dict] = {}
+        self.async_writers = []
+
+    def _sync_io(self):
+        for writer in self.async_writers:
+            writer.synchronize()
+            writer.fp.close()
+        self.async_writers.clear()
+
+    def _sync_d2h(self):
+        for writer in self.async_writers:
+            writer.sync_before_step()
+
+    def synchronize(self):
+        """This method must be called before updating the model weights."""
+        self._sync_d2h()
+
+    def __del__(self):
+        self._sync_d2h()
+        self._sync_io()
+
     def load_model(
         self, model: Union[nn.Module, ModelWrapper], checkpoint: str, strict: bool = True
     ) -> Union[nn.Module, ModelWrapper]:
@@ -111,6 +137,7 @@ def save_model(
         prefix: str = None,
         size_per_shard: int = 1024,
         use_safetensors: bool = False,
+        use_async: bool = False,
     ):
         """
         Save model to checkpoint.
@@ -138,11 +165,21 @@ def save_model(
             size_per_shard (int): size per shard in MB. Default: 1024. This value is only used when shard = True.
             use_safetensors (bool): whether to use safe tensors. Default: False. If set to True, the checkpoint will be saved
         """
+        self._sync_io()
+        if use_async and not use_safetensors:
+            logger = get_dist_logger()
+            logger.warning(
+                "Async save is only supported when use_safetensors is set to True. "
+                "Setting use_safetensors to True for async save."
+            )
+            use_safetensors = True
 
         if shard:
-            self.save_sharded_model(model, checkpoint, gather_dtensor, prefix, size_per_shard, use_safetensors)
+            self.save_sharded_model(
+                model, checkpoint, gather_dtensor, prefix, size_per_shard, use_safetensors, use_async=use_async
+            )
         else:
-            self.save_unsharded_model(model, checkpoint, gather_dtensor, use_safetensors)
+            self.save_unsharded_model(model, checkpoint, gather_dtensor, use_safetensors, use_async=use_async)
 
     def load_optimizer(self, optimizer: Optimizer, checkpoint: str, prefix: str = None, size_per_shard: int = 1024):
         """
@@ -234,6 +271,7 @@ def save_sharded_model(
         prefix: Optional[str],
         size_per_shard: int,
         use_safetensors: bool,
+        use_async: bool = False,
     ):
         """
         Save model to sharded checkpoint.
@@ -248,7 +286,9 @@ def save_sharded_model(
         """
 
     @abstractmethod
-    def save_unsharded_model(self, model: nn.Module, checkpoint: str, gather_dtensor: bool, use_safetensors: bool):
+    def save_unsharded_model(
+        self, model: nn.Module, checkpoint: str, gather_dtensor: bool, use_safetensors: bool, use_async: bool = False
+    ):
         """
         Save model to unsharded checkpoint.
 

colossalai/checkpoint_io/general_checkpoint_io.py

@@ -8,9 +8,13 @@
 import torch.nn as nn
 from torch.optim import Optimizer
 
+from colossalai.utils.safetensors import move_and_save
+
 from .checkpoint_io_base import CheckpointIO
 from .index_file import CheckpointIndexFile
 from .utils import (
+    async_save_state_dict_shards,
+    create_pinned_state_dict,
     get_model_base_filenames,
     get_optimizer_base_filenames,
     is_safetensors_available,
@@ -40,15 +44,27 @@ def load_unsharded_model(self, model: nn.Module, checkpoint: str, strict: bool):
         checkpoint = load_state_dict(checkpoint)
         model.load_state_dict(checkpoint, strict=strict)
 
-    def save_unsharded_model(self, model: nn.Module, checkpoint: str, gather_dtensor: bool, use_safetensors: bool):
+    def save_unsharded_model(
+        self, model: nn.Module, checkpoint: str, gather_dtensor: bool, use_safetensors: bool, use_async: bool = False
+    ):
         state_dict = model.state_dict()
 
         # TODO(FrankLeeeee): add support for gather_dtensor
         if gather_dtensor:
             pass
 
-        # save the checkpoint
-        save_state_dict(state_dict, checkpoint, use_safetensors)
+        if use_async:
+            from tensornvme.async_file_io import AsyncFileWriter
+
+            writer = AsyncFileWriter(open(checkpoint, "wb"), self.N_WRITE_ENTRIES, backend="pthread")
+            if id(model) not in self.pinned_state_dicts:
+                self.pinned_state_dicts[id(model)] = create_pinned_state_dict(state_dict)
+            self.async_writers.append(writer)
+            move_and_save(writer, state_dict, self.pinned_state_dicts[id(model)])
+        else:
+
+            # save the checkpoint
+            save_state_dict(state_dict, checkpoint, use_safetensors)
 
     def load_sharded_optimizer(self, optimizer: Optimizer, index_file_path: str, prefix: str):
         """
@@ -151,6 +167,7 @@ def save_sharded_model(
         prefix: Optional[str] = None,
         max_shard_size: int = 1024,
         use_safetensors: bool = False,
+        use_async: bool = False,
     ):
         """
         implement this method as it can be supported by Huggingface model,
@@ -168,16 +185,30 @@ def save_sharded_model(
         weights_name, save_index_file = get_model_base_filenames(prefix, use_safetensors)
         index_file = CheckpointIndexFile(checkpoint_path)
 
-        # Save shards of optimizer states.
-        # In general cases, is_master is set to True to get the right behavior.
-        total_size = save_state_dict_shards(
-            sharded_state_dict=state_dict_shard,
-            checkpoint=checkpoint_path,
-            index_file=index_file,
-            base_filename=weights_name,
-            is_master=True,
-            use_safetensors=use_safetensors,
-        )
+        if use_async:
+            pinned_state_dict = self.pinned_state_dicts.get(id(model), None)
+            total_size, new_pinned_state_dict, writers = async_save_state_dict_shards(
+                sharded_state_dict=state_dict_shard,
+                checkpoint=checkpoint_path,
+                index_file=index_file,
+                base_filename=weights_name,
+                is_master=True,
+                pinned_state_dict=pinned_state_dict,
+                n_write_entries=self.N_WRITE_ENTRIES,
+            )
+            self.pinned_state_dicts[id(model)] = new_pinned_state_dict
+            self.async_writers.extend(writers)
+        else:
+            # Save shards of optimizer states.
+            # In general cases, is_master is set to True to get the right behavior.
+            total_size = save_state_dict_shards(
+                sharded_state_dict=state_dict_shard,
+                checkpoint=checkpoint_path,
+                index_file=index_file,
+                base_filename=weights_name,
+                is_master=True,
+                use_safetensors=use_safetensors,
+            )
 
         index_file.append_meta_data("total_size", total_size)
         index_file.write_index_file(save_index_file)

colossalai/checkpoint_io/utils.py

@@ -5,7 +5,7 @@
 from collections import defaultdict
 from itertools import chain
 from pathlib import Path
-from typing import Iterator, List, Mapping, Optional, OrderedDict, Tuple
+from typing import Dict, Iterator, List, Mapping, Optional, OrderedDict, Tuple
 
 import torch
 import torch.nn as nn
@@ -19,6 +19,7 @@
     to_global,
     to_global_for_customized_distributed_tensor,
 )
+from colossalai.utils.safetensors import move_and_save
 
 SAFE_WEIGHTS_NAME = "model.safetensors"
 WEIGHTS_NAME = "pytorch_model.bin"
@@ -263,6 +264,71 @@ def save_state_dict_shards(
     return total_size
 
 
+def async_save_state_dict_shards(
+    sharded_state_dict: Iterator[Tuple[OrderedDict, int]],
+    checkpoint: str,
+    index_file: "CheckpointIndexFile",
+    base_filename: str,
+    is_master: bool,
+    pinned_state_dict: Optional[Dict[str, torch.Tensor]],
+    n_write_entries: int,
+    use_pp_format: bool = False,
+) -> Tuple[int, Dict[str, torch.Tensor], list]:
+    """
+    Save sharded state dict only on master rank, this method can be used by both model and optimizer states.
+    Args:
+        sharded_state_dict (Iterator[Tuple[OrderedDict, int]]): a generator of shards, each shard contains state dict and shard size.
+        checkpoint (str): The path of checkpoint directory as string.
+        index_file (CheckpointIndexFile): The index file object to be updated.
+        base_filename (str): Decides the prefix of filenames of shards.
+        is_master (bool): Whether current rank is main process.
+        use_safetensors (bool, optional): Whether to use safetensors to save checkpoint. Defaults to False.
+        use_pp_format: (bool, optional): Whether to save the files in pipeline format including stage information. Defaults to False.
+
+    Returns:
+        int: the total size of shards
+    """
+    from tensornvme.async_file_io import AsyncFileWriter
+
+    total_size = 0
+    shard_filenames = []
+    if pinned_state_dict is None:
+        returned_state_dict = {}
+    else:
+        returned_state_dict = pinned_state_dict
+    writers = []
+    for idx, shard_pair in enumerate(sharded_state_dict):
+        shard, current_size = shard_pair
+        # Just loop over the sharder and gather to other ranks if not master
+        if not is_master:
+            del shard
+            continue
+        shard_file = get_shard_filename(base_filename, idx)
+        total_size = total_size + current_size
+        for key in shard.keys():
+            index_file.append_weight_map(key, shard_file)
+        checkpoint_file_path = os.path.join(checkpoint, shard_file)
+
+        writer = AsyncFileWriter(open(checkpoint_file_path, "wb"), n_write_entries, backend="pthread")
+        writers.append(writer)
+
+        if pinned_state_dict is not None:
+            sub_pinned_state_dict = {k: pinned_state_dict[k] for k in shard.keys()}
+        else:
+            sub_pinned_state_dict = create_pinned_state_dict(shard)
+            returned_state_dict.update(sub_pinned_state_dict)
+
+        # Only save on master rank.
+        move_and_save(writer, shard, sub_pinned_state_dict)
+        shard_filenames.append(shard_file)
+        del shard
+
+    # Clean folder, deleted unneeded files.
+    clean_folder(checkpoint, base_filename, shard_filenames, is_master=is_master, use_pp_format=use_pp_format)
+
+    return total_size, returned_state_dict, writers
+
+
 def shard_model_checkpoint(state_dict: torch.Tensor, max_shard_size: int = 1024) -> Iterator[Tuple[OrderedDict, int]]:
     """
     Splits a model state dictionary in sub-checkpoints so that the final size of each sub-checkpoint does not exceed a
@@ -799,3 +865,10 @@ def get_shard_filename(weights_name: str, idx: int):
     shard_file = weights_name.replace(".bin", f"-{idx+1:05d}.bin")
     shard_file = shard_file.replace(".safetensors", f"-{idx+1:05d}.safetensors")
     return shard_file
+
+
+def create_pinned_state_dict(state_dict: Dict[str, torch.Tensor]):
+    pin_mem = dict()
+    for name, tensor in state_dict.items():
+        pin_mem[name] = torch.empty_like(tensor, pin_memory=True, device="cpu")
+    return pin_mem