OSError: [WinError 87] 参数错误

[liuziqian1]

mmdetection3.3.0 cuda 12.4 pytorch 2.5.0 训练时提示

Traceback (most recent call last):
File "D:\SOFT\miniconda\envs\open-mmlab\Lib\multiprocessing\queues.py", line 246, in _feed
send_bytes(obj)
File "D:\SOFT\miniconda\envs\open-mmlab\Lib\multiprocessing\connection.py", line 200, in send_bytes
self._send_bytes(m[offset:offset + size])
File "D:\SOFT\miniconda\envs\open-mmlab\Lib\multiprocessing\connection.py", line 289, in _send_bytes
ov, err = _winapi.WriteFile(self._handle, buf, overlapped=True)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
Traceback (most recent call last):
File "D:\python\mmdetection-main\tools\train.py", line 121, in
main()
File "D:\python\mmdetection-main\tools\train.py", line 117, in main
OSError: [WinError 87] 参数错误。

是为什么？

[liuziqian1]

System environment:
sys.platform: win32
Python: 3.11.11 | packaged by Anaconda, Inc. | (main, Dec 11 2024, 16:34:19) [MSC v.1929 64 bit (AMD64)]
CUDA available: True
MUSA available: False
numpy_random_seed: 400595520
GPU 0: NVIDIA GeForce RTX 3090
CUDA_HOME: C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.4
NVCC: Cuda compilation tools, release 12.4, V12.4.131
MSVC: 用于 x64 的 Microsoft (R) C/C++ 优化编译器 19.42.34436 版
GCC: n/a
PyTorch: 2.5.0
PyTorch compiling details: PyTorch built with:

• C++ Version: 201703

• MSVC 192930154

• Intel(R) oneAPI Math Kernel Library Version 2023.1-Product Build 20230303 for Intel(R) 64 architecture applications

• Intel(R) MKL-DNN v3.5.3 (Git Hash 66f0cb9eb66affd2da3bf5f8d897376f04aae6af)

• OpenMP 2019

• LAPACK is enabled (usually provided by MKL)

• CPU capability usage: AVX2

• CUDA Runtime 12.4

• NVCC architecture flags: -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

• CuDNN 90.1

• Magma 2.5.4

• Build settings: BLAS_INFO=mkl, BUILD_TYPE=Release, CUDA_VERSION=12.4, CUDNN_VERSION=9.1.0, CXX_COMPILER=C:/cb/pytorch_1000000000000/work/tmp_bin/sccache-cl.exe, CXX_FLAGS=/DWIN32 /D_WINDOWS /GR /EHsc /Zc:__cplusplus /bigobj /FS /utf-8 -DUSE_PTHREADPOOL -DNDEBUG -DUSE_KINETO -DLIBKINETO_NOCUPTI -DLIBKINETO_NOROCTRACER -DLIBKINETO_NOXPUPTI=ON -DUSE_FBGEMM -DUSE_XNNPACK -DSYMBOLICATE_MOBILE_DEBUG_HANDLE /wd4624 /wd4068 /wd4067 /wd4267 /wd4661 /wd4717 /wd4244 /wd4804 /wd4273, LAPACK_INFO=mkl, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, TORCH_VERSION=2.5.0, USE_CUDA=ON, USE_CUDNN=ON, USE_CUSPARSELT=OFF, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_GLOO=ON, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=OFF, USE_NNPACK=OFF, USE_OPENMP=ON, USE_ROCM=OFF, USE_ROCM_KERNEL_ASSERT=OFF,

  TorchVision: 0.20.0
  OpenCV: 4.11.0
  MMEngine: 0.10.6

Runtime environment:
cudnn_benchmark: False
mp_cfg: {'mp_start_method': 'fork', 'opencv_num_threads': 0}
dist_cfg: {'backend': 'nccl'}
seed: 400595520
Distributed launcher: none
Distributed training: False
GPU number: 1

[liuziqian1]

I use mmdetection3.3.0

[liuziqian1]

I use this : python D:\python\mmdetection-main\tools\train.py .\config\buildtest.py --work-dir work_dirs/cascade-mask-rcnn_x101-64x4d_fpn_20e_coco
my buildtest.py is :

新配置继承了基本配置，并做了必要的修改

base = '../cascade-mask-rcnn_x101-64x4d_fpn_20e_coco.py'

我们还需要更改 head 中的 num_classes 以匹配数据集中的类别数

model = dict(
roi_head=dict(
bbox_head=[dict(type='Shared2FCBBoxHead',num_classes=1),dict(type='Shared2FCBBoxHead',num_classes=1),dict(type='Shared2FCBBoxHead',num_classes=1)],
mask_head=dict(type='FCNMaskHead',num_classes=1)),
backbone=dict(init_cfg=None),
)

修改数据集相关配置

data_root = 'data/buildings/tiff/'
metainfo = {
'classes': ('building', ),
'palette': [
(220, 20, 60),
]
}
train_dataloader = dict(
batch_size=8,
dataset=dict(
data_root=data_root,
metainfo=metainfo,
ann_file='train/coco_annotations.json',
data_prefix=dict(img='train/')),
num_workers=2)

train_pipeline = [

dict(backend_args=None, type='LoadImageFromFile'),

dict(type='LoadAnnotations', with_bbox=True, with_mask=True),

dict(keep_ratio=True, scale=( # 会对图片进行缩放 ，keep_ratio 是true表示会等比例缩放 scale 表示图片会缩放到的大小

1333, 800

#512 , 512

), type='Resize'),

dict(prob=0.5, type='RandomFlip'),

dict(type='PackDetInputs'),

]

val_dataloader = dict(
batch_size=8, #默认就是1
dataset=dict(
data_root=data_root,
metainfo=metainfo,
ann_file='val/coco_annotations.json',
data_prefix=dict(img='val/')),
num_workers=2)

val_evaluator = dict(

dataset=dict(

ann_file='data/balloon/val/val.json',

)

type='CocoMetric',

metric=['bbox', 'segm'],

format_only=False)

test_dataloader = dict(
batch_size=8, #默认就是1
dataset=dict(
data_root=data_root,
metainfo=metainfo,
ann_file='test/coco_annotations.json',
data_prefix=dict(img='test/')),
num_workers=2)

optim_wrapper = dict(
type='OptimWrapper',
optimizer=dict(type='SGD', lr=0.002, momentum=0.9, weight_decay=0.0001))

修改评价指标相关配置

val_evaluator = dict(ann_file=data_root + 'val/coco_annotations.json')
test_evaluator = dict(ann_file=data_root + 'test/coco_annotations.json')

log_config = dict(

interval=10, # 每10个iter记录一次

hooks=[

dict(type='TextLoggerHook'), # 文本日志

dict(type='TensorboardLoggerHook'), # TensorBoard 日志

])

使用预训练的 Mask R-CNN 模型权重来做初始化，可以提高模型性能

load_from = 'cascade_mask_rcnn_x101_64x4d_fpn_20e_coco_20200512_161033-bdb5126a.pth'