The package mmcv won't be able to install

[JiahaoYu1]

Hello,

I'm a student on stage in Prof. Jeffery Cardille's lab at McGill, and we're trying to run this AI model on our own computer. When I was trying out the Flood Segmentation, the package mmcv is required, and I tried different ways to install the package but always get the error: Failed building wheel for mmcv, and here are my torch and cuda version:

!python -c 'import torch;print(torch.__version__);print(torch.version.cuda)'
2.2.1+cu121
12.1

and some approaches that I tried:
!pip install -U openmim

1. !mim install mmcv==2.0.0 -f https://download.openmmlab.com/mmcv/dist/cu121/torch2.2.1/index.html

2. !mim install mmcv-full

3. !pip install mmcv==2.1.0 -f https://download.openmmlab.com/mmcv/dist/cu121/torch2.1/index.html

4. git clone https://github.com/open-mmlab/mmcv.git
cd mmcv
pip install -r requirements/optional.txt
MMCV_WITH_OPS=1 pip install -e .
python .dev_scripts/check_installation.py

Thank you!

[melgor]

I aslo was trying to run the model, but with the success. Here is my path:

conda create -n ibm python==3.10
conda activate ibm
pip3 install torch torchvision
pip install -U openmim
git clone https://github.com/NASA-IMPACT/hls-foundation-os.git
cd hls-foundation-os/
pip install -e .
mim install mmcv-full

Before running pip install -e . I changed setup.py from repo, exactly removed all versions of required python libraries, except mmsegmentation

from setuptools import setup

setup(
    name="geospatial_fm",
    version="0.1.0",
    description="MMSegmentation classes for geospatial-fm finetuning",
    author="Paolo Fraccaro, Carlos Gomes, Johannes Jakubik",
    packages=["geospatial_fm"],
    license="Apache 2",
    long_description=open("README.md").read(),
    install_requires=[
        "mmsegmentation @ git+https://github.com/open-mmlab/mmsegmentation.git@186572a3ce64ac9b6b37e66d58c76515000c3280",
        "rasterio",
        "rioxarray",
        "einops",
        "timm",
        "tensorboard",
        "imagecodecs",
        "yapf",
    ],
)

It was working like a charm!
Note: installing mmcv-full can take even 30 minutes!

[JiahaoYu1]

Hi,

Right now the mmcv installation part works for me, but I'm running into another issue where when I'm trying to run this line of code: finetuned_model = init_model(Config.fromfile(config_path), ckpt, device="cpu"), I'm getting errors like ImportError: cannot import name 'print_log' from 'mmcv' , and ImportError: Failed to import custom modules from {'imports': ['geospatial_fm']},

also: ImportError Traceback (most recent call last) [/usr/local/lib/python3.10/dist-packages/mmengine/utils/misc.py](https://localhost:8080/#) in import_modules_from_strings(imports, allow_failed_imports) 76 try: ---> 77 imported_tmp = import_module(imp) 78 except ImportError:

I'm assuming this is also related to the mmcv import issue? Thanks