[Bug] MMDistributedDataParallel distributed training can not help save memory. The total memory usage is twice that of a single card.
humian321 opened this issue · 2 comments
Prerequisite
- I have searched Issues and Discussions but cannot get the expected help.
- The bug has not been fixed in the latest version(https://github.com/open-mmlab/mmengine).
Environment
OrderedDict([('sys.platform', 'linux'), ('Python', '3.8.18 (default, Sep 11 2023, 13:40:15) [GCC 11.2.0]'), ('CUDA available', True), ('MUSA available', False), ('numpy_random_seed', 2147483648), ('GPU 0,1', 'NVIDIA GeForce RTX 3090'), ('CUDA_HOME', '/usr/local/cuda'), ('NVCC', 'Cuda compilation tools, release 11.8, V11.8.89'), ('GCC', 'gcc (Ubuntu 9.4.0-1ubuntu1~20.04.1) 9.4.0'), ('PyTorch', '2.0.1'), ('PyTorch compiling details', 'PyTorch built with:\n - GCC 9.3\n - C++ Version: 201703\n - Intel(R) oneAPI Math Kernel Library Version 2023.1-Product Build 20230303 for Intel(R) 64 architecture applications\n - Intel(R) MKL-DNN v2.7.3 (Git Hash 6dbeffbae1f23cbbeae17adb7b5b13f1f37c080e)\n - OpenMP 201511 (a.k.a. OpenMP 4.5)\n - LAPACK is enabled (usually provided by MKL)\n - NNPACK is enabled\n - CPU capability usage: AVX2\n - CUDA Runtime 11.8\n - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_61,code=sm_61;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75;-gencode;arch=compute_80,code=sm_80;-gencode;arch=compute_86,code=sm_86;-gencode;arch=compute_90,code=sm_90;-gencode;arch=compute_37,code=compute_37\n - CuDNN 8.7\n - Magma 2.6.1\n - Build settings: BLAS_INFO=mkl, BUILD_TYPE=Release, CUDA_VERSION=11.8, CUDNN_VERSION=8.7.0, CXX_COMPILER=/opt/rh/devtoolset-9/root/usr/bin/c++, CXX_FLAGS= -D_GLIBCXX_USE_CXX11_ABI=0 -fabi-version=11 -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -DNDEBUG -DUSE_KINETO -DLIBKINETO_NOROCTRACER -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DSYMBOLICATE_MOBILE_DEBUG_HANDLE -O2 -fPIC -Wall -Wextra -Werror=return-type -Werror=non-virtual-dtor -Werror=bool-operation -Wnarrowing -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wunused-local-typedefs -Wno-unused-parameter -Wno-unused-function -Wno-unused-result -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Werror=cast-function-type -Wno-stringop-overflow, LAPACK_INFO=mkl, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, TORCH_DISABLE_GPU_ASSERTS=ON, TORCH_VERSION=2.0.1, USE_CUDA=ON, USE_CUDNN=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, USE_ROCM=OFF, \n'), ('TorchVision', '0.15.2'), ('OpenCV', '4.9.0'), ('MMEngine', '0.10.3')])
Reproduces the problem - code sample
cfg.py:
model_wrapper_cfg=dict(
type='MMDistributedDataParallel',
find_unused_parameters=True,
)
train.py
def parse_args():
parser = argparse.ArgumentParser(description="Train a segmentor")
parser.add_argument("config", help="train config file path")
parser.add_argument('--checkpoint', default="none", help='checkpoint file')
parser.add_argument("--work-dir", help="the dir to save logs and models")
parser.add_argument(
"--resume",
action="store_true",
default=False,
help="resume from the latest checkpoint in the work_dir automatically",
)
parser.add_argument(
"--amp",
action="store_true",
default=False,
help="enable automatic-mixed-precision training",
)
parser.add_argument(
"--cfg-options",
nargs="+",
action=DictAction,
help="override some settings in the used config, the key-value pair "
"in xxx=yyy format will be merged into config file. If the value to "
'be overwritten is a list, it should be like key="[a,b]" or key=a,b '
'It also allows nested list/tuple values, e.g. key="[(a,b),(c,d)]" '
"Note that the quotation marks are necessary and that no white space "
"is allowed.",
)
parser.add_argument(
"--launcher",
choices=["none", "pytorch", "slurm", "mpi"],
default="none",
help="job launcher",
)
# When using PyTorch version >= 2.0.0, the `torch.distributed.launch`
# will pass the `--local-rank` parameter to `tools/train.py` instead
# of `--local_rank`.
parser.add_argument("--local_rank", "--local-rank", type=int, default=0)
args = parser.parse_args()
if "LOCAL_RANK" not in os.environ:
os.environ["LOCAL_RANK"] = str(args.local_rank)
return args
def main():
args = parse_args()
# load config
cfg = Config.fromfile(args.config)
cfg.launcher = args.launcher
if args.cfg_options is not None:
cfg.merge_from_dict(args.cfg_options)
# work_dir is determined in this priority: CLI > segment in file > filename
if args.work_dir is not None:
# update configs according to CLI args if args.work_dir is not None
cfg.work_dir = args.work_dir
elif cfg.get("work_dir", None) is None:
# use config filename as default work_dir if cfg.work_dir is None
cfg.work_dir = osp.join(
"./work_dirs", osp.splitext(osp.basename(args.config))[0]
)
# enable automatic-mixed-precision training
if args.amp is True:
optim_wrapper = cfg.optim_wrapper.type
if optim_wrapper == "AmpOptimWrapper":
print_log(
"AMP training is already enabled in your config.",
logger="current",
level=logging.WARNING,
)
else:
assert optim_wrapper == "OptimWrapper", (
"`--amp` is only supported when the optimizer wrapper type is "
f"`OptimWrapper` but got {optim_wrapper}."
)
cfg.optim_wrapper.type = "AmpOptimWrapper"
cfg.optim_wrapper.loss_scale = "dynamic"
# resume training
cfg.resume = args.resume
runner = Runner.from_cfg(cfg)
runner.train()
if __name__ == "__main__":
main()
Reproduces the problem - command or script
python -m torch.distributed.launch --nnode=1 --node_rank=0 --nproc_per_node=2 tools/train.py configs/dinov2/rein_dinov2_mask2former_512x512_bs1x4.py --launcher pytorch
Reproduces the problem - error message
using two cards:
Thu Feb 22 22:58:08 2024
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 525.105.17 Driver Version: 525.105.17 CUDA Version: 12.0 |
|-------------------------------+----------------------+----------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|===============================+======================+======================|
| 0 NVIDIA GeForce ... On | 00000000:4F:00.0 Off | N/A |
| 32% 53C P2 287W / 350W | 24084MiB / 24576MiB | 53% Default |
| | | N/A |
+-------------------------------+----------------------+----------------------+
| 1 NVIDIA GeForce ... On | 00000000:D5:00.0 Off | N/A |
| 32% 53C P2 276W / 350W | 24168MiB / 24576MiB | 53% Default |
| | | N/A |
+-------------------------------+----------------------+----------------------+
+-----------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=============================================================================|
+-----------------------------------------------------------------------------+
using on card:
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 525.105.17 Driver Version: 525.105.17 CUDA Version: 12.0 |
|-------------------------------+----------------------+----------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|===============================+======================+======================|
| 0 NVIDIA GeForce ... On | 00000000:4F:00.0 Off | N/A |
| 33% 56C P2 312W / 350W | 23026MiB / 24576MiB | 100% Default |
| | | N/A |
+-------------------------------+----------------------+----------------------+
| 1 NVIDIA GeForce ... On | 00000000:D5:00.0 Off | N/A |
| 33% 38C P8 21W / 350W | 2MiB / 24576MiB | 0% Default |
| | | N/A |
+-------------------------------+----------------------+----------------------+
+-----------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=============================================================================|
+-----------------------------------------------------------------------------+
Additional information
No response
Since the batch_size of each of the two GPUs is the same as the batch_size of the original GPU, there is no drop in memory. The total number of samples consumed per iteration for both GPUs is 2 * batch size, so you can see that the training time is nearly halved.
@zhouzaida Thanks for your reply, I understand now.