"Cannot pickle 'module' object error " when train my own dataset using mmdetection.

[kdh4672]

Describe the bug

I'm now trying to train my own dataset.
when I run
(mmd2) daehyeon@daehyeon-MS-7C13:~/mmdetection$ python tools/train.py configs/mask_rcnn_r50_fpn_1x.py ,

I got error like

yapf:enable
runtime settings
total_epochs = 8
dist_params = dict(backend='nccl')
log_level = 'INFO'
work_dir = './work_dirs/mask_rcnn_r50_fpn_1x_test'
load_from = None
resume_from = None
workflow = [('train', 1)]

loading annotations into memory...
Done (t=0.00s)
creating index...
index created!
2020-04-11 06:20:11,015 - mmdet - INFO - Start running, host: daehyeon@daehyeon-MS-7C13, work_dir: /home/daehyeon/mmdetection/work_dirs/mask_rcnn_r50_fpn_1x_test
2020-04-11 06:20:11,015 - mmdet - INFO - workflow: [('train', 1)], max: 8 epochs
Traceback (most recent call last):
File "tools/train.py", line 144, in
main()
File "tools/train.py", line 132, in main
train_detector(
File "/home/daehyeon/anaconda3/envs/mmd2/lib/python3.8/site-packages/mmdet-1.1.0+51df8a9-py3.8-linux-x86_64.egg/mmdet/apis/train.py", line 104, in train_detector
_non_dist_train(
File "/home/daehyeon/anaconda3/envs/mmd2/lib/python3.8/site-packages/mmdet-1.1.0+51df8a9-py3.8-linux-x86_64.egg/mmdet/apis/train.py", line 225, in _non_dist_train
runner.run(data_loaders, cfg.workflow, cfg.total_epochs)
File "/home/daehyeon/anaconda3/envs/mmd2/lib/python3.8/site-packages/mmcv/runner/runner.py", line 359, in run
epoch_runner(data_loaders[i], **kwargs)
File "/home/daehyeon/anaconda3/envs/mmd2/lib/python3.8/site-packages/mmcv/runner/runner.py", line 273, in train
self.call_hook('after_train_epoch')
File "/home/daehyeon/anaconda3/envs/mmd2/lib/python3.8/site-packages/mmcv/runner/runner.py", line 226, in call_hook
getattr(hook, fn_name)(self)
File "/home/daehyeon/anaconda3/envs/mmd2/lib/python3.8/site-packages/mmcv/runner/dist_utils.py", line 74, in wrapper
return func(*args, **kwargs)
File "/home/daehyeon/anaconda3/envs/mmd2/lib/python3.8/site-packages/mmcv/runner/hooks/checkpoint.py", line 26, in after_train_epoch
runner.save_checkpoint(
File "/home/daehyeon/anaconda3/envs/mmd2/lib/python3.8/site-packages/mmcv/runner/runner.py", line 247, in save_checkpoint
save_checkpoint(self.model, filepath, optimizer=optimizer, meta=meta)
File "/home/daehyeon/anaconda3/envs/mmd2/lib/python3.8/site-packages/mmcv/runner/checkpoint.py", line 249, in save_checkpoint
torch.save(checkpoint, filename)
File "/home/daehyeon/anaconda3/envs/mmd2/lib/python3.8/site-packages/torch/serialization.py", line 327, in save
_legacy_save(obj, opened_file, pickle_module, pickle_protocol)
File "/home/daehyeon/anaconda3/envs/mmd2/lib/python3.8/site-packages/torch/serialization.py", line 400, in _legacy_save
pickler.dump(obj)
TypeError: cannot pickle 'module' object

I don't know what ' cannot pickle 'module' object' does mean and what should I do.

+) but when i run same thing
(mmd2) daehyeon@daehyeon-MS-7C13:~/mmdetection$ python tools/train.py configs/mask_rcnn_r50_fpn_1x.py ,

but using COCO dataset (not my own dataset), It works perfectly like

loading annotations into memory...
Done (t=0.39s)
creating index...
index created!
2020-04-11 06:45:36,680 - mmdet - INFO - Start running, host: daehyeon@daehyeon-MS-7C13, work_dir: /home/daehyeon/mmdetection/work_dirs/mask_rcnn_r50_fpn_1x
2020-04-11 06:45:36,680 - mmdet - INFO - workflow: [('train', 1)], max: 12 epochs
2020-04-11 06:46:15,204 - mmdet - INFO - Epoch [1][50/2476] lr: 0.00797, eta: 6:20:47, time: 0.770, data_time: 0.026, memory: 3828, loss_rpn_cls: 0.3816, loss_rpn_bbox: 0.0893, loss_cls: 0.6315, acc: 92.6523, loss_bbox: 0.1025, loss_mask: 0.7153, loss: 1.9201
2020-04-11 06:46:54,318 - mmdet - INFO - Epoch [1][100/2476] lr: 0.00931, eta: 6:23:06, time: 0.782, data_time: 0.014, memory: 3828, loss_rpn_cls: 0.2121, loss_rpn_bbox: 0.0620, loss_cls: 0.4719, acc: 93.7617, loss_bbox: 0.1379, loss_mask: 0.6908, loss: 1.5747

and in this case, data_type = 'MyDataset' , data_type = 'CocoDataset' both does work

Reproduction
I read getstart.md for use myowndataset
and i did

1. At mmdetection/mmdet/my_dataset.py, I wrote

from .coco import CocoDataset
from .registry import DATASETS

@DATASETS.register_module

class MyDataset(CocoDataset):

CLASSES = ('lactofit', 'vitamin')

2. At mmdetection/mmdet/init.py I wrote add

from .my_dataset import MyDataset

all = [
'CustomDataset', 'XMLDataset', 'CocoDataset', 'VOCDataset',
'CityscapesDataset', 'GroupSampler', 'DistributedGroupSampler',
'build_dataloader', 'ConcatDataset', 'RepeatDataset', 'WIDERFaceDataset',
'DATASETS', 'build_dataset','MyDataset'
]

3. At configs/mask_rcnn_r50_fpn_1x.py i edit
   1)) dataset_type = 'MyDataset'

2)) bbox_head=dict(num_classes= 3) ## because my own dataset has 2 classes. and 1 is for
## background
3)) mask_head=dict(num_classes=3)

4))
data_root = '../pycococreator/examples/shapes/train/' ## it is my data root

data = dict( imgs_per_gpu=2,
workers_per_gpu=2,
train=dict(type=dataset_type,
ann_file=data_root + 'instances_nutrients_train.json',
img_prefix=data_root + 'images_test/',
pipeline=train_pipeline),

val=dict(type=dataset_type,
    ann_file=data_root + 'instances_nutrients_train.json',
    img_prefix=data_root + 'images_test/',
    pipeline=test_pipeline),

test=dict(
    type=dataset_type,
    ann_file=data_root + 'instances_nutrients_train.json',
    img_prefix=data_root + 'images_test/',
    pipeline=test_pipeline))

4. At mmdetection/mmdet/core/evaluation/class_names.py

i edit

def coco_classes():
return [
'lactofit', 'vitamin'
]

Finally) (mmd2) daehyeon@daehyeon-MS-7C13:~/mmdetection$ python setup.py develop

This is my dataset link:
https://github.com/kdh4672/my_own_dataset_nutrients_mmdetection.git

i made coco format json file by using pycococreator from 👍 https://github.com/waspinator/pycococreator

Environment
when i get my environment information i got some trouble, but I solved.
-->
My mmdet's version is '1.1.0+51df8a9' and mmdet.version.py has version = '1.1.0+51df8a9' collect_env.py doesn't work because of mmdet.version.py like
" AttributeError: module 'mmdet' has no attribute 'version' "
so i just edit this line in collect_env.py
env_info['MMDetection'] = mmdet.__version -> env_info['MMDetection'] = '1.1.0+51df8a9'

then collect_env works:
(mmd2) daehyeon@daehyeon-MS-7C13:/mmdetection$ python mmdet/utils/collect_env.py
sys.platform: linux
Python: 3.8.2 (default, Mar 26 2020, 15:53:00) [GCC 7.3.0]
CUDA available: True
CUDA_HOME: /usr/local/cuda-10.1
NVCC: Cuda compilation tools, release 10.1, V10.1.105
GPU 0: GeForce GTX 1660
GCC: gcc (Ubuntu 7.5.0-3ubuntu118.04) 7.5.0
PyTorch: 1.4.0
PyTorch compiling details: PyTorch built with:

• GCC 7.3
• Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
• Intel(R) MKL-DNN v0.21.1 (Git Hash 7d2fd500bc78936d1d648ca713b901012f470dbc)
• OpenMP 201511 (a.k.a. OpenMP 4.5)
• NNPACK is enabled
• CUDA Runtime 10.1
• 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_37,code=compute_37
• CuDNN 7.6.3
• Magma 2.5.1
• Build settings: BLAS=MKL, BUILD_NAMEDTENSOR=OFF, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -fopenmp -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -O2 -fPIC -Wno-narrowing -Wall -Wextra -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -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 -Wno-stringop-overflow, DISABLE_NUMA=1, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=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_STATIC_DISPATCH=OFF,

TorchVision: 0.5.0
OpenCV: 4.2.0
MMCV: 0.4.2
MMDetection: 1.1.0+51df8a9
MMDetection Compiler: GCC 7.4
MMDetection CUDA Compiler: 9.1

Error traceback

Traceback (most recent call last):
File "tools/train.py", line 144, in
main()
File "tools/train.py", line 132, in main
train_detector(
File "/home/daehyeon/anaconda3/envs/mmd2/lib/python3.8/site-packages/mmdet-1.1.0+51df8a9-py3.8-linux-x86_64.egg/mmdet/apis/train.py", line 104, in train_detector
_non_dist_train(
File "/home/daehyeon/anaconda3/envs/mmd2/lib/python3.8/site-packages/mmdet-1.1.0+51df8a9-py3.8-linux-x86_64.egg/mmdet/apis/train.py", line 225, in _non_dist_train
runner.run(data_loaders, cfg.workflow, cfg.total_epochs)
File "/home/daehyeon/anaconda3/envs/mmd2/lib/python3.8/site-packages/mmcv/runner/runner.py", line 359, in run
epoch_runner(data_loaders[i], **kwargs)
File "/home/daehyeon/anaconda3/envs/mmd2/lib/python3.8/site-packages/mmcv/runner/runner.py", line 273, in train
self.call_hook('after_train_epoch')
File "/home/daehyeon/anaconda3/envs/mmd2/lib/python3.8/site-packages/mmcv/runner/runner.py", line 226, in call_hook
getattr(hook, fn_name)(self)
File "/home/daehyeon/anaconda3/envs/mmd2/lib/python3.8/site-packages/mmcv/runner/dist_utils.py", line 74, in wrapper
return func(*args, **kwargs)
File "/home/daehyeon/anaconda3/envs/mmd2/lib/python3.8/site-packages/mmcv/runner/hooks/checkpoint.py", line 26, in after_train_epoch
runner.save_checkpoint(
File "/home/daehyeon/anaconda3/envs/mmd2/lib/python3.8/site-packages/mmcv/runner/runner.py", line 247, in save_checkpoint
save_checkpoint(self.model, filepath, optimizer=optimizer, meta=meta)
File "/home/daehyeon/anaconda3/envs/mmd2/lib/python3.8/site-packages/mmcv/runner/checkpoint.py", line 249, in save_checkpoint
torch.save(checkpoint, filename)
File "/home/daehyeon/anaconda3/envs/mmd2/lib/python3.8/site-packages/torch/serialization.py", line 327, in save
_legacy_save(obj, opened_file, pickle_module, pickle_protocol)
File "/home/daehyeon/anaconda3/envs/mmd2/lib/python3.8/site-packages/torch/serialization.py", line 400, in _legacy_save
pickler.dump(obj)
TypeError: cannot pickle 'module' object

[kdh4672]

If someone has same problem, I recommend using Dectectron2.
I start using Detectron2 2 days ago, now I can train my dataset.
They have detail documentation and more easy.

[kdh4672]

https://github.com/facebookresearch/detectron2 it s a object detection tool like mmdetection -----Original Message----- From: "striver6"<notifications@github.com> To: "open-mmlab/mmdetection"<mmdetection@noreply.github.com>; Cc: "kdh4672"<kdh4672@naver.com>; "Mention"<mention@noreply.github.com>; Sent: 2020-05-17 (일) 12:24:49 (GMT+09:00) Subject: Re: [open-mmlab/mmdetection] "Cannot pickle 'module' object error " when train my own dataset using mmdetection. (#2425) @kdh4672 What's Dectectron2 ? I have met the same issus. please give more detials. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or unsubscribe.