Open jere1882 opened 1 year ago
I solved this one by installing timm 0.3.0 - Anyways, after this, there were loads of issues to deal with. The repo doesn't really work as published here, it requires a lot of grooming. Go use a different segmentation ViT unless you really need to use this one.
I have the same problem, and i solved it as follows:
# file environment_open_mmlab.yaml is at the back
conda env create -f environment_open_mmlab.yaml
conda activate open-mmlab
cd SETR
pip install -e . --user
pip install numpy==1.23.5
And be careful, you need to change file mmseg/models/decode_heads/hlg_utils.py
as follows.
from
# from
19 from mmseg.models.backbones.hlg_share import AttentionShareGlobal, AttentionShareLocal, DWMlp, InnerPatchEmbed
# to
19 from mmseg.models.backbones.hlg import AttentionShareGlobal, AttentionShareLocal, DWMlp, InnerPatchEmbed
The file environment_open_mmlab.yaml
mentioned above is as follows.
name: open-mmlab
channels:
- pytorch
- defaults
- conda-forge
dependencies:
- _libgcc_mutex=0.1=main
- _openmp_mutex=4.5=1_gnu
- blas=1.0=mkl
- ca-certificates=2021.7.5=h06a4308_1
- certifi=2021.5.30=py39h06a4308_0
- cudatoolkit=10.1.243=h6bb024c_0
- freetype=2.10.4=h5ab3b9f_0
- intel-openmp=2021.3.0=h06a4308_3350
- jpeg=9b=h024ee3a_2
- lcms2=2.12=h3be6417_0
- ld_impl_linux-64=2.35.1=h7274673_9
- libffi=3.3=he6710b0_2
- libgcc-ng=9.3.0=h5101ec6_17
- libgomp=9.3.0=h5101ec6_17
- libpng=1.6.37=hbc83047_0
- libstdcxx-ng=9.3.0=hd4cf53a_17
- libtiff=4.2.0=h85742a9_0
- libwebp-base=1.2.0=h27cfd23_0
- lz4-c=1.9.3=h295c915_1
- mkl=2021.3.0=h06a4308_520
- mkl-service=2.4.0=py39h7f8727e_0
- mkl_fft=1.3.0=py39h42c9631_2
- mkl_random=1.2.2=py39h51133e4_0
- ncurses=6.2=he6710b0_1
- ninja=1.10.2=hff7bd54_1
- numpy=1.20.3=py39hf144106_0
- numpy-base=1.20.3=py39h74d4b33_0
- olefile=0.46=pyhd3eb1b0_0
- openjpeg=2.4.0=h3ad879b_0
- openssl=1.1.1l=h7f8727e_0
- pillow=8.3.1=py39h2c7a002_0
- pip=21.2.4=py39h06a4308_0
- python=3.9.7=h12debd9_1
- python_abi=3.9=2_cp39
- pytorch=1.7.1=py3.9_cuda10.1.243_cudnn7.6.3_0
- readline=8.1=h27cfd23_0
- setuptools=58.0.4=py39h06a4308_0
- six=1.16.0=pyhd3eb1b0_0
- sqlite=3.36.0=hc218d9a_0
- tk=8.6.11=h1ccaba5_0
- torchaudio=0.7.2=py39
- torchvision=0.8.2=py39_cu101
- typing_extensions=3.10.0.2=pyh06a4308_0
- tzdata=2021a=h5d7bf9c_0
- wheel=0.37.0=pyhd3eb1b0_1
- xz=5.2.5=h7b6447c_0
- zlib=1.2.11=h7b6447c_3
- zstd=1.4.9=haebb681_0
- pip:
- addict==2.4.0
- attr==0.3.1
- backcall==0.2.0
- decorator==5.1.0
- ipython==7.28.0
- jedi==0.18.0
- kornia==0.5.11
- matplotlib-inline==0.1.3
- mmcv-full==1.2.7
- opencv-python==4.5.1.48
- packaging==21.0
- parso==0.8.2
- pexpect==4.8.0
- pickleshare==0.7.5
- prompt-toolkit==3.0.20
- ptyprocess==0.7.0
- pygments==2.10.0
- pyyaml==5.4.1
- timm==0.3.2
- traitlets==5.1.0
- wcwidth==0.2.5
- yapf==0.31.0
I just followed the exact steps indicated in the ubuntu section of install.md
I cannot run training tough, because I get:
ModuleNotFoundError: No module named 'timm'
Simply running
pip install timm
messes up everything because it upgrades to torch 1.13.0. What do I do?