mmcv-full version not satisfied according to the given url in Dockerfile

[luciagan]

I built the environment from the given Dockerfile with mmcv-full part as belows

RUN pip install mmcv-full==latest+torch1.6.0+cu101 -f https://download.openmmlab.com/mmcv/dist/index.html

but the latest version of mmcv-full is 1.3.5 according to the given url https://download.openmmlab.com/mmcv/dist/index.html, which causes the following ERROR when start training:

`*****************************************
Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed.

---

Traceback (most recent call last):
File "./tools/train.py", line 13, in
from mmseg import version
File "/mmsegmentation/mmseg/init.py", line 59, in
f'MMCV=={mmcv.version} is used but incompatible. '
AssertionError: MMCV==1.3.5 is used but incompatible. Please install mmcv>=(1, 3, 7, 0, 0, 0), <=(1, 4, 0, 0, 0, 0).
Traceback (most recent call last):
File "./tools/train.py", line 13, in
from mmseg import version
File "/mmsegmentation/mmseg/init.py", line 59, in
f'MMCV=={mmcv.version} is used but incompatible. '
AssertionError: MMCV==1.3.5 is used but incompatible. Please install mmcv>=(1, 3, 7, 0, 0, 0), <=(1, 4, 0, 0, 0, 0).
Traceback (most recent call last):
File "./tools/train.py", line 13, in
from mmseg import version
File "/mmsegmentation/mmseg/init.py", line 59, in
f'MMCV=={mmcv.version} is used but incompatible. '
AssertionError: MMCV==1.3.5 is used but incompatible. Please install mmcv>=(1, 3, 7, 0, 0, 0), <=(1, 4, 0, 0, 0, 0).
Traceback (most recent call last):
File "./tools/train.py", line 13, in
from mmseg import version
File "/mmsegmentation/mmseg/init.py", line 59, in
f'MMCV=={mmcv.version} is used but incompatible. '
AssertionError: MMCV==1.3.5 is used but incompatible. Please install mmcv>=(1, 3, 7, 0, 0, 0), <=(1, 4, 0, 0, 0, 0).
Traceback (most recent call last):
File "/opt/conda/lib/python3.7/runpy.py", line 193, in _run_module_as_main
"main", mod_spec)
File "/opt/conda/lib/python3.7/runpy.py", line 85, in _run_code
exec(code, run_globals)
File "/opt/conda/lib/python3.7/site-packages/torch/distributed/launch.py", line 261, in
main()
File "/opt/conda/lib/python3.7/site-packages/torch/distributed/launch.py", line 257, in main
cmd=cmd)
subprocess.CalledProcessError: Command '['/opt/conda/bin/python', '-u', './tools/train.py', '--local_rank=3', './configs/setr/setr_pup_512x512_160k_b16_ade20k_jslseg.py', '--launcher', 'pytorch', '--load-from', 'checkpoints/setr/pretrained/setr_pup_512x512_160k_b16_ade20k_20210619_191343-7e0ce826.pth']' returned non-zero exit status 1.
`

[MengzhangLI]

Hi, based on your description. I think you should use MMCV-full version >=1.3.7 and <1.4.0, which would be compatibal with latest MMSegmentation.

I can not open the link you mention "latest version==1.3.5". But anyways, you can follow the instruction here for solving it. I will check out dockerfile on my own.

Please let us know whether you fix it after MMCV-full installation.

Best,

[luciagan]

Hi, @MengzhangLI it works following the mmcv instruction you suggested. I install the correct mmcv through pip install mmcv-full -f https://download.openmmlab.com/mmcv/dist/{cu_version}/{torch_version}/index.html

Thanks a lot for your timely reply.

Best,

[luciagan]

Hi, @MengzhangLI maybe it is still necessary to check the Dockerfile anyway. It seems that the given url in the Dockerfile does not match mmsegmentation-v0.17.0 when installing mmcv, because the latest version of mmcv in this given url is 1.3.5 while it requires mmcv>=(1, 3, 7, 0, 0, 0), <=(1, 4, 0, 0, 0, 0) using v0.17.0.

[MengzhangLI]

OK, I will write it down and discuss it with my colleagues.

Thanks for your nice reminder.

Best,

[luciagan]

Thanks a lot. Wish you all the best! @MengzhangLI

[Junjun2016]

Hi @luciagan
FIx #860
Have a try.

[luciagan]

Hi, @Junjun2016 thanks!