[Bug] mim resources not found: [...]/.venv/lib/python3.10/site-packages/mmpretrain/.mim, you can try to install the latest mmpretrain.

[maygeisler]

Branch

main branch (mmpretrain version)

Describe the bug

When I try to clone mmpretrain from source (cloning the github repo) and install it with mim mim install -e .
i get mim resources not found: [...]/.venv/lib/python3.10/site-packages/mmpretrain/.mim, you can try to install the latest mmpretrain. as a warning.

When i then try to run the demo python demo/image_demo.py demo/demo.JPEG resnet18_8xb32_in1k --device cpu i get OSError: File '[...]/.venv/lib/python3.10/site-packages/mmpretrain/.mim/model-index.yml' does not exist.

I tried to create said path and copy the model-index.yml found in the repo there and try to run the demo again i get ValueError: Unavailable model "resnet18_8xb32_in1k", you can specify find a model name or a config file or find a model name from https://mmpretrain.readthedocs.io/en/latest/modelzoo_statistics.html#all-checkpoints

I cant get mmpretrain up and running at all.

Steps taken:

1. Create python venv
2. install pytorch pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118
3. install openmim pip install -U openmim
4. install mmcv and mmengine with openmim mim install mmcv mim install mmengine
5. clone repo git clone https://github.com/open-mmlab/mmpretrain.git
6. cd into it cd mmpretrain
7. run mim install -e .

Environment

{'sys.platform': 'linux',
'Python': '3.10.12 (main, Nov 20 2023, 15:14:05) [GCC 11.4.0]',
'CUDA available': True,
'MUSA available': False,
'numpy_random_seed': 2147483648,
'GPU 0': 'NVIDIA GeForce RTX 4080',
'CUDA_HOME': '/usr/local/cuda-11.8',
'NVCC': 'Cuda compilation tools, release 11.8, V11.8.89',
'GCC': 'x86_64-linux-gnu-gcc (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0',
'PyTorch': '2.2.1+cu118',
'TorchVision': '0.17.1+cu118',
'OpenCV': '4.9.0',
'MMEngine': '0.10.3',
'MMCV': '2.0.1',
'MMPreTrain': '1.0.1+'}

Other information

I did not modify the code in any way.

I tried the newest version from the repo and the latest release (1.0.1).

I also tried to install the latest version from mim mim install "mmpretrain>=1.0.0rc8" which gives me the same warning.

I'm using a virtual environment.
I installed pytorch via pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118
i also installed mim pip install -U openmim and mmcv and mmengine with mim mim install mmcv mim install mmengine