ImportError: cannot import name 'MultiScaleDeformableAttention' from 'mmcv.cnn.bricks.transformer'

FrancescoSaverioZuppichini commented 3 years ago

Thanks for reporting the unexpected results and we appreciate it a lot.

Checklist

I have searched related issues but cannot get the expected help.
I have read the FAQ documentation but cannot get the expected help.
The unexpected results still exist in the latest version.

Describe the Issue

/home/zuppif/integration-object-detection-icevision/.venv/lib/python3.9/site-packages/mmcv/cnn/bricks/transformer.py:28: UserWarning: Fail to import ``MultiScaleDeformableAttention`` from ``mmcv.ops.multi_scale_deform_attn``, You should install ``mmcv-full`` if you need this module. 
  warnings.warn('Fail to import ``MultiScaleDeformableAttention`` from '
/home/zuppif/integration-object-detection-icevision/.venv/lib/python3.9/site-packages/mmdet/models/utils/transformer.py:26: UserWarning: `MultiScaleDeformableAttention` in MMCV has been moved to `mmcv.ops.multi_scale_deform_attn`, please update your MMCV
  warnings.warn(
Traceback (most recent call last):
  File "/home/zuppif/integration-object-detection-icevision/.venv/lib/python3.9/site-packages/mmdet/models/utils/transformer.py", line 23, in <module>
    from mmcv.ops.multi_scale_deform_attn import MultiScaleDeformableAttention
  File "/home/zuppif/integration-object-detection-icevision/.venv/lib/python3.9/site-packages/mmcv/ops/__init__.py", line 2, in <module>
    from .ball_query import ball_query
  File "/home/zuppif/integration-object-detection-icevision/.venv/lib/python3.9/site-packages/mmcv/ops/ball_query.py", line 7, in <module>
    ext_module = ext_loader.load_ext('_ext', ['ball_query_forward'])
  File "/home/zuppif/integration-object-detection-icevision/.venv/lib/python3.9/site-packages/mmcv/utils/ext_loader.py", line 13, in load_ext
    ext = importlib.import_module('mmcv.' + name)
  File "/home/zuppif/miniconda3/lib/python3.9/importlib/__init__.py", line 127, in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
ImportError: /home/zuppif/integration-object-detection-icevision/.venv/lib/python3.9/site-packages/mmcv/_ext.cpython-39-x86_64-linux-gnu.so: undefined symbol: _ZNK2at6Tensor6deviceEv

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/home/zuppif/integration-object-detection-icevision/notebook.py", line 7, in <module>
    from icevision import ObjectDetectionRecord, Parser, ClassMap, BBox, ImgSize
  File "/home/zuppif/integration-object-detection-icevision/.venv/lib/python3.9/site-packages/icevision/__init__.py", line 7, in <module>
    from icevision import models
  File "/home/zuppif/integration-object-detection-icevision/.venv/lib/python3.9/site-packages/icevision/models/__init__.py", line 15, in <module>
    from icevision.models import mmdet
  File "/home/zuppif/integration-object-detection-icevision/.venv/lib/python3.9/site-packages/icevision/models/mmdet/__init__.py", line 1, in <module>
    from icevision.models.mmdet import common
  File "/home/zuppif/integration-object-detection-icevision/.venv/lib/python3.9/site-packages/icevision/models/mmdet/common/__init__.py", line 3, in <module>
    from icevision.models.mmdet.common import bbox
  File "/home/zuppif/integration-object-detection-icevision/.venv/lib/python3.9/site-packages/icevision/models/mmdet/common/bbox/__init__.py", line 1, in <module>
    from icevision.models.mmdet.common.bbox.dataloaders import *
  File "/home/zuppif/integration-object-detection-icevision/.venv/lib/python3.9/site-packages/icevision/models/mmdet/common/bbox/dataloaders.py", line 10, in <module>
    from icevision.models.mmdet.common.utils import convert_background_from_zero_to_last
  File "/home/zuppif/integration-object-detection-icevision/.venv/lib/python3.9/site-packages/icevision/models/mmdet/common/utils.py", line 11, in <module>
    from mmdet.models import build_detector
  File "/home/zuppif/integration-object-detection-icevision/.venv/lib/python3.9/site-packages/mmdet/models/__init__.py", line 2, in <module>
    from .backbones import *  # noqa: F401,F403
  File "/home/zuppif/integration-object-detection-icevision/.venv/lib/python3.9/site-packages/mmdet/models/backbones/__init__.py", line 2, in <module>
    from .csp_darknet import CSPDarknet
  File "/home/zuppif/integration-object-detection-icevision/.venv/lib/python3.9/site-packages/mmdet/models/backbones/csp_darknet.py", line 11, in <module>
    from ..utils import CSPLayer
  File "/home/zuppif/integration-object-detection-icevision/.venv/lib/python3.9/site-packages/mmdet/models/utils/__init__.py", line 16, in <module>
    from .transformer import (DetrTransformerDecoder, DetrTransformerDecoderLayer,
  File "/home/zuppif/integration-object-detection-icevision/.venv/lib/python3.9/site-packages/mmdet/models/utils/transformer.py", line 29, in <module>
    from mmcv.cnn.bricks.transformer import MultiScaleDeformableAttention
ImportError: cannot import name 'MultiScaleDeformableAttention' from 'mmcv.cnn.bricks.transformer' (/home/zuppif/integration-object-detection-icevision/.venv/lib/python3.9/site-packages/mmcv/cnn/bricks/transformer.py)

Reproduction

Install mmcv-full using mim or

pip install mmcv-full -f https://download.openmmlab.com/mmcv/dist/cu102/torch1.9.1/index.html

then try to use vfnet

Environment

Please run python -c "from mmcv.utils import collect_env; print(collect_env())" to collect necessary environment information \ It results in an error

python -c "from mmcv.utils import collect_env; print(collect_env())"
Traceback (most recent call last):
File "<string>", line 1, in <module>
File "/home/zuppif/integration-object-detection-icevision/.venv/lib/python3.9/site-packages/mmcv/utils/env.py", line 87, in collect_env
from mmcv.ops import get_compiler_version, get_compiling_cuda_version
File "/home/zuppif/integration-object-detection-icevision/.venv/lib/python3.9/site-packages/mmcv/ops/__init__.py", line 2, in <module>
from .ball_query import ball_query
File "/home/zuppif/integration-object-detection-icevision/.venv/lib/python3.9/site-packages/mmcv/ops/ball_query.py", line 7, in <module>
ext_module = ext_loader.load_ext('_ext', ['ball_query_forward'])
File "/home/zuppif/integration-object-detection-icevision/.venv/lib/python3.9/site-packages/mmcv/utils/ext_loader.py", line 13, in load_ext
ext = importlib.import_module('mmcv.' + name)
File "/home/zuppif/miniconda3/lib/python3.9/importlib/__init__.py", line 127, in import_module
return _bootstrap._gcd_import(name[level:], package, level)
ImportError: /home/zuppif/integration-object-detection-icevision/.venv/lib/python3.9/site-packages/mmcv/_ext.cpython-39-x86_64-linux-gnu.so: undefined symbol: _ZNK2at6Tensor6deviceEv

zhouzaida commented 3 years ago

hi, had the mmcv-full been installed? please provide the output of pip list | grep mmcv-full

zhouzaida commented 3 years ago

hi, @FrancescoSaverioZuppichini , in the latest version (v1.3.15), we provide pre-compiled packages for the Python 3.9 environment. You can install the latest mmcv-full with the following command.

pip install mmcv-full -f https://download.openmmlab.com/mmcv/dist/cu102/torch1.9.0/index.html

FrancescoSaverioZuppichini commented 3 years ago

Thank you very much, I'll check this out

zhouzaida commented 3 years ago

hi, is there any progress?

FrancescoSaverioZuppichini commented 3 years ago

yes so let me share a minimum example container in which the problem arises

FROM nvcr.io/nvidia/cuda:11.2.0-cudnn8-runtime-ubuntu20.04

RUN apt-get -yq update && \
    DEBIAN_FRONTEND=noninteractive apt-get install -yqq --no-install-recommends \
    git \
    ssh \
    wget \
    python3-dev \ 
    python3-pip \ 
    python3-opencv \ 
    build-essential \
    gcc \
    && \
    apt-get clean && \
    rm -rf /var/lib/apt/lists/*

RUN pip install cython
RUN pip install mmcv-full==1.3.16 -f https://download.openmmlab.com/mmcv/dist/cu112/torch1.9.1/index.html
RUN pip install mmdet==2.17.0

env

{'CUDA available': True,
 'CUDA_HOME': '/usr/local/cuda',
 'GCC': 'gcc (Ubuntu 9.3.0-17ubuntu1~20.04) 9.3.0',
 'GPU 0,1,2': 'GeForce GTX 1080 Ti',
 'MMCV': '1.3.16',
 'MMCV CUDA Compiler': 'n/a',
 'MMCV Compiler': 'n/a',
 'NVCC': '',
 'OpenCV': '4.5.4-dev',
 'PyTorch': '1.9.1+cu102',
 'PyTorch compiling details': 'PyTorch built with:\n'
                              '  - GCC 7.3\n'
                              '  - C++ Version: 201402\n'
                              '  - Intel(R) Math Kernel Library Version '
                              '2020.0.0 Product Build 20191122 for Intel(R) 64 '
                              'architecture applications\n'
                              '  - Intel(R) MKL-DNN v2.1.2 (Git Hash '
                              '98be7e8afa711dc9b66c8ff3504129cb82013cdb)\n'
                              '  - OpenMP 201511 (a.k.a. OpenMP 4.5)\n'
                              '  - NNPACK is enabled\n'
                              '  - CPU capability usage: AVX2\n'
                              '  - CUDA Runtime 10.2\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_70,code=sm_70\n'
                              '  - CuDNN 7.6.5\n'
                              '  - Magma 2.5.2\n'
                              '  - Build settings: BLAS_INFO=mkl, '
                              'BUILD_TYPE=Release, CUDA_VERSION=10.2, '
                              'CUDNN_VERSION=7.6.5, '
                              'CXX_COMPILER=/opt/rh/devtoolset-7/root/usr/bin/c++, '
                              'CXX_FLAGS= -Wno-deprecated '
                              '-fvisibility-inlines-hidden -DUSE_PTHREADPOOL '
                              '-fopenmp -DNDEBUG -DUSE_KINETO -DUSE_FBGEMM '
                              '-DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK '
                              '-DUSE_XNNPACK -DSYMBOLICATE_MOBILE_DEBUG_HANDLE '
                              '-O2 -fPIC -Wno-narrowing -Wall -Wextra '
                              '-Werror=return-type '
                              '-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-unused-local-typedefs -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 '
                              '-Wno-stringop-overflow, LAPACK_INFO=mkl, '
                              'PERF_WITH_AVX=1, PERF_WITH_AVX2=1, '
                              'PERF_WITH_AVX512=1, TORCH_VERSION=1.9.1, '
                              '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, \n',
 'Python': '3.8.10 (default, Sep 28 2021, 16:10:42) [GCC 9.3.0]',
 'TorchVision': '0.10.1+cu102',
 'sys.platform': 'linux'}

zhouzaida commented 3 years ago

| pip install mmcv-full==1.3.16 -f https://download.openmmlab.com/mmcv/dist/cu112/torch1.9.1/index.html

Hi, we do not provide pre-compiled packages for cu112 and torch1.9.1, but we do provide cu111 and torch1.9.0.

You can use the follow command to install it.

pip install mmcv-full==1.3.16 -f https://download.openmmlab.com/mmcv/dist/cu111/torch1.9.0/index.html

dwight-foster commented 3 years ago

Hello I am getting the same error. I used pip install mmcv-full==1.3.16 -f https://download.openmmlab.com/mmcv/dist/cu111/torch1.9.0/index.html, but it has not changed anything. The full error is

`/opt/conda/lib/python3.7/site-packages/mmdet/models/utils/transformer.py:27: UserWarning: MultiScaleDeformableAttention` in MMCV has been moved to `mmcv.ops.multi_scale_deform_attn`, please update your MMCV '`MultiScaleDeformableAttention` in MMCV has been moved to '

ImportError Traceback (most recent call last) /opt/conda/lib/python3.7/site-packages/mmdet/models/utils/transformer.py in 22 try: ---> 23 from mmcv.ops.multi_scale_deform_attn import MultiScaleDeformableAttention 24

/opt/conda/lib/python3.7/site-packages/mmcv/ops/init.py in 1 # Copyright (c) OpenMMLab. All rights reserved. ----> 2 from .assign_score_withk import assign_score_withk 3 from .ball_query import ball_query

/opt/conda/lib/python3.7/site-packages/mmcv/ops/assign_score_withk.py in 5 ext_module = ext_loader.load_ext( ----> 6 '_ext', ['assign_score_withk_forward', 'assign_score_withk_backward']) 7

/opt/conda/lib/python3.7/site-packages/mmcv/utils/ext_loader.py in load_ext(name, funcs) 12 def load_ext(name, funcs): ---> 13 ext = importlib.import_module('mmcv.' + name) 14 for fun in funcs:

/opt/conda/lib/python3.7/importlib/init.py in import_module(name, package) 126 level += 1 --> 127 return _bootstrap._gcd_import(name[level:], package, level) 128

ImportError: libtorch_cpu.so: cannot open shared object file: No such file or directory

During handling of the above exception, another exception occurred:

ImportError Traceback (most recent call last)

in ----> 1 from mmdet.models import build_detector 2 from config import get_config 3 from mmcv import Config, DictAction 4 5 args = parse_args() /opt/conda/lib/python3.7/site-packages/mmdet/models/__init__.py in 1 # Copyright (c) OpenMMLab. All rights reserved. ----> 2 from .backbones import * # noqa: F401,F403 3 from .builder import (BACKBONES, DETECTORS, HEADS, LOSSES, NECKS, 4 ROI_EXTRACTORS, SHARED_HEADS, build_backbone, 5 build_detector, build_head, build_loss, build_neck, /opt/conda/lib/python3.7/site-packages/mmdet/models/backbones/__init__.py in 1 # Copyright (c) OpenMMLab. All rights reserved. ----> 2 from .csp_darknet import CSPDarknet 3 from .darknet import Darknet 4 from .detectors_resnet import DetectoRS_ResNet 5 from .detectors_resnext import DetectoRS_ResNeXt /opt/conda/lib/python3.7/site-packages/mmdet/models/backbones/csp_darknet.py in 9 10 from ..builder import BACKBONES ---> 11 from ..utils import CSPLayer 12 13 /opt/conda/lib/python3.7/site-packages/mmdet/models/utils/__init__.py in 14 from .res_layer import ResLayer, SimplifiedBasicBlock 15 from .se_layer import SELayer ---> 16 from .transformer import (DetrTransformerDecoder, DetrTransformerDecoderLayer, 17 DynamicConv, PatchEmbed, Transformer, nchw_to_nlc, 18 nlc_to_nchw) /opt/conda/lib/python3.7/site-packages/mmdet/models/utils/transformer.py in 27 '`MultiScaleDeformableAttention` in MMCV has been moved to ' 28 '`mmcv.ops.multi_scale_deform_attn`, please update your MMCV') ---> 29 from mmcv.cnn.bricks.transformer import MultiScaleDeformableAttention 30 31 ImportError: cannot import name 'MultiScaleDeformableAttention' from 'mmcv.cnn.bricks.transformer' (/opt/conda/lib/python3.7/site-packages/mmcv/cnn/bricks/transformer.py)` I cannot seem to figure out why this error keeps coming back. The correct version is installed I checked it. Can anyone help?

deepwilson commented 2 years ago

Same issue. Looking forward to a solution!

tabmoo commented 2 years ago

Me too!

amine0110 commented 2 years ago

Same thing here!

zhouzaida commented 2 years ago

Hi, what is your mmcv version?

amine0110 commented 2 years ago

@zhouzaida it is 1.5 the same as mmcv-full

I checked if the paths and files are put in the right places, and everything is put without any problem!

And when I tried to do the import without the try statement, I got this:

ImportError: DLL load failed: The specified module could not be found.

Matesanz commented 2 years ago

The problem is that ImportError in this line in is silently catching another exception rather than the "install mmcv-full" one.

if you try the line that is raising the Error:

from mmcv.ops.multi_scale_deform_attn import MultiScaleDeformableAttention
>>> ImportError: .../lib/python3.9/site-packages/mmcv/_ext.cpython-39-x86_64-linux-gnu.so: undefined symbol: _ZNK2at6Tensor4add_ERKS0_RKN3c106ScalarE

Solution

For me was install mmcv CPU version:

pip install mmcv-full -f https://download.openmmlab.com/mmcv/dist/cpu/torch1.11.0/index.html

qzhai commented 2 years ago

Hi, what is your mmcv version?

Same question.

I test the mmcv==1.6.0, cu113, pytorch=1.9.0, and the same question when using the CPU version.

ganesh3 commented 2 years ago

same question when I create a docker container with the below commands RUN pip install mmcv-full==1.3.7 -f https://download.openmmlab.com/mmcv/dist/cu101/torch1.8.0/index.html -U RUN pip install mmdet==2.13.0 -U RUN pip install mmsegmentation==0.17.0 -U

howardchina commented 1 year ago

same question, why always the same question?

terminal: python3 -m pip list | grep mmcv-full mmcv-full 1.7.0

error: "/home/data/programs/miniconda3/envs/openmmlab/lib/python3.8/site-packages/mmcv/c nn/bricks/transformer.py:33: UserWarning: Fail to import MultiScaleDeformableA ttention from mmcv.ops.multi_scale_deform_attn, You should install mmcv- full if you need this module."

AlistairMcCutcheon commented 1 year ago

I am running pytorch 1.8 and cuda 11.1 in a docker container.

I had the issue when running (this installed mmcv-full 1.7.0): RUN pip install openmim RUN mim install mmcv-full

However this version of mmcv-full fixed the issue without needing to change to a CPU version: RUN pip install mmcv-full -f https://download.openmmlab.com/mmcv/dist/cu111/torch1.8.0/mmcv_full-1.7.1-cp39-cp39-manylinux1_x86_64.whl

anamazingclown commented 8 months ago

I also had the issue when running my code.

environments:

torch：1.9.0

torchvision：0.10.0

mmcv-full==1.4.7

mmdet==2.20.0

GPU 0,1: NVIDIA GeForce RTX 4090,

CUDA Version: 12.1

The way I installed mmcv-full on the anaconda virtual environment on the far server is: pip install mmcv_full-1.4.7-cp37-cp37m-manylinux1_x86_64.whl

18780165950 commented 3 months ago

Me too, does anyone solve this problem？

open-mmlab / mmcv