[checkpointio] support load-pin overlap (#6177)

* [checkpointio] support load-pin overlap

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

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

* [test] add conftest

---------

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

colossalai/booster/plugin/gemini_plugin.py

@@ -20,7 +20,7 @@
     create_pinned_state_dict,
     get_model_base_filenames,
     get_optimizer_base_filenames,
-    load_shard_state_dict,
+    load_state_dict_shards,
     save_config_file,
     save_state_dict,
     save_state_dict_shards,
@@ -29,7 +29,6 @@
 from colossalai.interface import ModelWrapper, OptimizerWrapper
 from colossalai.logging import get_dist_logger
 from colossalai.shardformer import ShardConfig, ShardFormer
-from colossalai.utils.safetensors import load_flat
 from colossalai.zero import GeminiDDP, GeminiOptimizer
 from colossalai.zero.gemini.memory_tracer import MemStats
 
@@ -350,11 +349,9 @@ def load_sharded_optimizer(
 
         # Load optimizer states from shard files under checkpoint path.
         # For each file, only load the states managed by current process.
-        for shard_file in checkpoint_files:
-            if shard_file.endswith(".safetensors"):
-                state_dict_shard = load_flat(shard_file)
-            else:
-                state_dict_shard = load_shard_state_dict(Path(shard_file), use_safetensors=False)
+        for state_dict_shard in load_state_dict_shards(
+            checkpoint_files, True, False, low_cpu_mem_mode=low_cpu_mem_mode
+        ):
             if not low_cpu_mem_mode:
                 state_dict_shard = create_pinned_state_dict(state_dict_shard, empty=False, num_threads=num_threads)
             optimizer.load_param_states(state_dict_shard)

colossalai/booster/plugin/low_level_zero_plugin.py

@@ -24,8 +24,8 @@
     get_optimizer_base_filenames,
     get_shard_filename,
     load_param_groups_into_optimizer,
-    load_shard_state_dict,
     load_state_dict,
+    load_state_dict_shards,
     load_states_into_optimizer,
     save_param_groups,
     save_state_dict,
@@ -276,13 +276,7 @@ def load_sharded_optimizer(
 
         checkpoint_files, _ = ckpt_index_file.get_checkpoint_filenames()
 
-        for shard_file in checkpoint_files:
-            if shard_file.endswith(".safetensors"):
-                from colossalai.utils.safetensors import load_flat
-
-                state_dict = load_flat(shard_file)
-            else:
-                state_dict = load_shard_state_dict(Path(shard_file), use_safetensors=False)
+        for state_dict in load_state_dict_shards(checkpoint_files, True, False, low_cpu_mem_mode):
             # shard state dict
             for param_idx, state in state_dict.items():
                 for k, v in state.items():

colossalai/booster/plugin/torch_fsdp_plugin.py

@@ -255,8 +255,8 @@ def load_sharded_model(
         checkpoint_files, _ = ckpt_index_file.get_checkpoint_filenames()
 
         fsdp_state_dict = {}
-        for shard_file in checkpoint_files:
-            fsdp_state_dict.update(utils.load_shard_state_dict(Path(shard_file), use_safetensors))
+        for state_dict in utils.load_state_dict_shards(checkpoint_files, False, use_safetensors):
+            fsdp_state_dict.update(state_dict)
 
         with FSDP.state_dict_type(model.unwrap(), StateDictType.FULL_STATE_DICT):
             model.unwrap().load_state_dict(fsdp_state_dict, strict=False)
@@ -388,11 +388,7 @@ def load_sharded_optimizer(
         # Load param
         fsdp_optim_state = {}
         checkpoint_files, _ = ckpt_index_file.get_checkpoint_filenames()
-        for shard_file in checkpoint_files:
-            if shard_file.endswith(".safetensors"):
-                state_dict_shard = load_flat(shard_file, seperator=".")
-            else:
-                state_dict_shard = utils.load_shard_state_dict(Path(shard_file), use_safetensors=False)
+        for state_dict_shard in utils.load_state_dict_shards(checkpoint_files, True, False):
             fsdp_optim_state.update(state_dict_shard)
 
         fsdp_optim_dict = dict(state=fsdp_optim_state, param_groups=saved_param_groups)

colossalai/checkpoint_io/general_checkpoint_io.py

@@ -18,9 +18,9 @@
     get_optimizer_base_filenames,
     is_safetensors_available,
     load_param_groups_into_optimizer,
-    load_shard_state_dict,
     load_state_dict,
     load_state_dict_into_model,
+    load_state_dict_shards,
     load_states_into_optimizer,
     save_config_file,
     save_param_groups,
@@ -94,11 +94,7 @@ def load_sharded_optimizer(
 
         checkpoint_files, _ = ckpt_index_file.get_checkpoint_filenames()
 
-        for shard_file in checkpoint_files:
-            if shard_file.endswith(".safetensors"):
-                state_dict = load_flat(shard_file)
-            else:
-                state_dict = load_shard_state_dict(Path(shard_file), use_safetensors=False)
+        for state_dict in load_state_dict_shards(checkpoint_files, True, False, low_cpu_mem_mode):
             if not low_cpu_mem_mode:
                 state_dict = create_pinned_state_dict(state_dict, empty=False, num_threads=num_threads)
             load_states_into_optimizer(optimizer, state_dict, id_map)
@@ -295,8 +291,7 @@ def load_sharded_model(
         checkpoint_files, _ = ckpt_index_file.get_checkpoint_filenames()
         missing_keys = []
 
-        for shard_file in checkpoint_files:
-            state_dict = load_shard_state_dict(Path(shard_file), use_safetensors)
+        for state_dict in load_state_dict_shards(checkpoint_files, False, use_safetensors, low_cpu_mem_mode):
             if not low_cpu_mem_mode:
                 state_dict = create_pinned_state_dict(state_dict, empty=False, num_threads=num_threads)
             load_state_dict_into_model(model, state_dict, missing_keys, strict, load_sub_module)

colossalai/checkpoint_io/utils.py

@@ -6,7 +6,7 @@
 from collections import defaultdict
 from itertools import chain
 from pathlib import Path
-from typing import Dict, Iterator, List, Mapping, Optional, OrderedDict, Tuple, Union
+from typing import Dict, Generator, Iterator, List, Mapping, Optional, OrderedDict, Tuple, Union
 
 import torch
 import torch.nn as nn
@@ -21,7 +21,7 @@
     to_global,
     to_global_for_customized_distributed_tensor,
 )
-from colossalai.utils.safetensors import _flatten_optim_state_dict
+from colossalai.utils.safetensors import _flatten_optim_state_dict, load_flat
 
 SAFE_WEIGHTS_NAME = "model.safetensors"
 WEIGHTS_NAME = "pytorch_model.bin"
@@ -972,3 +972,35 @@ def create_pinned_state_dict(
                 idx = future_to_idx[future]
                 elems[idx] = future.result()
             return tree_unflatten(elems, spec)
+
+
+def load_optim_or_model_shard(path: str, is_optim: bool, use_safetensors: bool) -> dict:
+    if is_optim:
+        if path.endswith(".safetensors"):
+            state_dict = load_flat(path)
+        else:
+            state_dict = load_shard_state_dict(Path(path), use_safetensors=False)
+    else:
+        state_dict = load_shard_state_dict(Path(path), use_safetensors)
+    return state_dict
+
+
+def load_state_dict_shards(
+    checkpoint_files: List[str],
+    is_optim: bool,
+    use_safetensors: bool,
+    low_cpu_mem_mode: bool = True,
+    prefetch: int = 3,
+) -> Generator[dict, None, None]:
+    if low_cpu_mem_mode:
+        for shard_file in checkpoint_files:
+            state_dict = load_optim_or_model_shard(shard_file, is_optim, use_safetensors)
+            yield state_dict
+    else:
+        with concurrent.futures.ThreadPoolExecutor(max_workers=prefetch) as executor:
+            futures = []
+            for shard_file in checkpoint_files:
+                future = executor.submit(load_optim_or_model_shard, shard_file, is_optim, use_safetensors)
+                futures.append(future)
+            for future in concurrent.futures.as_completed(futures):
+                yield future.result()

tests/conftest.py

@@ -0,0 +1,10 @@
+import gc
+
+from colossalai.accelerator import get_accelerator
+
+
+def pytest_runtest_setup(item):
+    # called for running each test in 'a' directory
+    accelerator = get_accelerator()
+    accelerator.empty_cache()
+    gc.collect()