github/ColossalAI
1
0
Fork 0
mirror of https://github.com/hpcaitech/ColossalAI.git synced 2025-09-18 16:00:49 +00:00
Code Issues Packages Projects Releases Wiki Activity

[zero] add tensor placement policies (#743)

Browse Source 
* add tensor placement policies

* polish comments

* polish comments

* update moe unit tests
This commit is contained in:
ver217
2022-04-13 15:00:48 +08:00
committed by GitHub
parent 22c4b88d56
commit e396bb71f2
11 changed files with 139 additions and 83 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
tests/test_zero/common.py
View File

@@ -13,14 +13,12 @@ MP_PARALLEL_CONFIG = dict(fp16=dict(mode=None,), parallel=dict(pipeline=dict(siz
_ZERO_MODEL_CONFIG = dict(reduce_scatter_bucket_size_mb=25,
fp32_reduce_scatter=False,
offload_config=None,
tensor_placement_policy='cuda',
gradient_predivide_factor=1.0,
use_memory_tracer=False,
shard_strategy=TensorShardStrategy(),
reuse_fp16_shard=False)
_ZERO_OPTIMIZER_CONFIG = dict(cpu_offload=False,
initial_scale=2**5,
_ZERO_OPTIMIZER_CONFIG = dict(initial_scale=2**5,
min_scale=1,
growth_factor=2,
backoff_factor=0.5,

8
tests/test_zero/test_found_inf.py
View File

@@ -37,16 +37,12 @@ def _run_test_found_inf(cpu_offload, shard_strategy_class, gpu_margin_mem_ratio)
zero_model = ShardedModelV2(
zero_model,
shard_strategy,
offload_config=dict(device='cpu') if cpu_offload else None,
use_memory_tracer=gpu_margin_mem_ratio > 0.0,
tensor_placement_policy='cpu' if cpu_offload else 'cuda',
reuse_fp16_shard=True,
)
sharded_optim = HybridAdam(zero_model.parameters(), lr=1e-3)
sharded_optim = ShardedOptimizerV2(zero_model,
sharded_optim,
cpu_offload=cpu_offload,
gpu_margin_mem_ratio=gpu_margin_mem_ratio)
sharded_optim = ShardedOptimizerV2(zero_model, sharded_optim, gpu_margin_mem_ratio=gpu_margin_mem_ratio)
for i, (data, label) in enumerate(train_dataloader):
if i > 1:

2
tests/test_zero/test_shard_model_v2.py
View File

@@ -33,7 +33,7 @@ def run_model_test(enable_autocast, shard_strategy_class):
shard_strategy=shard_strategy,
shard_param=True):
zero_model = model_builder(checkpoint=True)
zero_model = ShardedModelV2(zero_model, shard_strategy, use_memory_tracer=True)
zero_model = ShardedModelV2(zero_model, shard_strategy)
model = model_builder(checkpoint=True).half()
col_model_deepcopy(zero_model, model)

4
tests/test_zero/test_sharded_optim_v2.py
View File

@@ -64,8 +64,7 @@ def _run_test_sharded_optim_v2(cpu_offload, shard_strategy_class, use_cpuadam, g
zero_model = ShardedModelV2(
zero_model,
shard_strategy,
offload_config=dict(device='cpu') if cpu_offload else None,
use_memory_tracer=gpu_margin_mem_ratio > 0.0,
tensor_placement_policy='cpu' if cpu_offload else 'cuda',
reuse_fp16_shard=use_cpuadam,
)
@@ -79,7 +78,6 @@ def _run_test_sharded_optim_v2(cpu_offload, shard_strategy_class, use_cpuadam, g
sharded_optim = optimizer_class(zero_model.parameters(), lr=1e-3)
sharded_optim = ShardedOptimizerV2(zero_model,
sharded_optim,
cpu_offload=cpu_offload,
initial_scale=2**5,
gpu_margin_mem_ratio=gpu_margin_mem_ratio)

4
tests/test_zero/test_stateful_tensor_mgr.py
View File

@@ -14,6 +14,7 @@ from colossalai.testing import rerun_on_exception
from torch.nn.parameter import Parameter
from typing import List
from functools import partial
from colossalai.zero.utils.tensor_placement_policy import AutoTensorPlacementPolicy
class Net(torch.nn.Module):
@@ -37,7 +38,8 @@ def run_stm():
p.colo_attr = ShardedParamV2(p, set_data_none=True)
GLOBAL_MODEL_DATA_TRACER.register_model(model)
mem_collector = MemStatsCollector()
stateful_tensor_mgr = StatefulTensorMgr(mem_collector)
tensor_placement_policy = AutoTensorPlacementPolicy(mem_stats_collector=mem_collector)
stateful_tensor_mgr = StatefulTensorMgr(tensor_placement_policy)
for p in model.parameters():
stateful_tensor_mgr.register_stateful_param(p.colo_attr)