vim-mamba-setup.text

### Image Set Upsudo apt-get install -y nvidia-docker2

### 1- Nvidia driver:

sudo apt install nvidia-driver-440
reboot
nvidia-smi ## verify installation

### 2- docker engine : https://docs.docker.com/engine/install/ubuntu/

- Don't use docker desktop,use the terminal only
- Make sure you're using the default context and not desktop-linux : 

docker context use default

### 3- Nvidia container toolkit :

curl -s -L https://nvidia.github.io/nvidia-docker/gpgkey | sudo apt-key add -
curl -s -L https://nvidia.github.io/nvidia-docker/ubuntu18.04/nvidia-docker.list | sudo tee /etc/apt/sources.list.d/nvidia-docker.list
sudo apt-get update
sudo apt-get install -y nvidia-container-toolkit
sudo apt-get install -y nvidia-docker2
sudo systemctl restart docker

### 4- Test that nvidia driver is recognized with a small image: 

docker run --name=test --gpus all -it ubuntu # if this doesn't work try reinstalling docker : sudo apt-get install --reinstall docker-ce 
nvidia-smi # it should work if it works you can exit the bash and stop the container.

### 5- Pull image with cuda11.8 :

docker pull nvidia/cuda:11.8.0-cudnn8-devel-ubuntu22.04
docker run --volume=/home/abdelnour/Documents/projects/Breast-Cancer-Classification:/home/Breast-Cancer-Classification --ipc=host --name=vim --gpus all -it nvidia/cuda:11.8.0-cudnn8-devel-ubuntu22.04
nvidia-smi

### 6- Python3 & pip :

apt update 
apt install python3
python3 --version # make sure >=3.9
install pip3
install pip

### 7- PyTorch :

pip3 install torch==2.1.1 torchvision==0.16.1 torchaudio==2.1.1 --index-url https://download.pytorch.org/whl/cu118
python3 -c "import torch;print(torch.__version__);print(torch.cuda.is_available())" # Output should be : 2.1.1+cu118,True

### 8- Setup repo :

apt install git
git clone https://github.com/hustvl/Vim.git
cd Vim
git checkout 6143d07b3dd31f904c63840a19e22d95d1124493
pip install -r vim/vim_requirements.txt
pip install -e mamba-1p1p1
pip show causal_conv1d # If the version is not 1.1.0 the repo won't work
pip uninstall causal_conv1d
pip install causal_conv1d==1.1.0
pip show causal_conv1d # Version should be 1.1.0

### 9- Test the repo
cd ..
apt install nano
nano test.py

- the content :

"""
import torch
import timm

print(torch.__version__)
print(torch.cuda.is_available())

from Vim.vim.models_mamba import (
    vim_tiny_patch16_224_bimambav2_final_pool_mean_abs_pos_embed_with_midclstok_div2 as vim_tiny_patch16_224
)

model = timm.create_model("vim_tiny_patch16_224_bimambav2_final_pool_mean_abs_pos_embed_with_midclstok_div2").cuda()

x = torch.randn(32, 3, 224, 224).cuda()

y = model(x)

print(y.shape)
"""

python3 test.py # Output should be : 2.1.1+cu118,True,torch.Size([32, 1000])