配环境常用指令.txt

cuda:
conda install -c nvidia/label/cuda-11.8.0 cuda-toolkit=11.8  锁11.8所有的相关依赖
conda install -c nvidia cuda-toolkit=11.2  会装一些不相关的依赖，不推荐
conda install cuda=12.1.0 -c nvidia  会装一些不相关的依赖，不推荐
conda list | grep cuda

查看资源使用情况：
lscpu 
top
df -h 看硬盘
free -h 看内存
nvidia-smi
gpustat

用ssh服务器的时候记得选择Interpreter不然vscode会用默认的

Pytorch:
conda install pytorch==1.13.1 torchvision==0.14.1 torchaudio==0.13.1 pytorch-cuda=11.7 -c pytorch -c nvidia
pip install --upgrade charset-normalizer requests
pip show torch

移除环境：
conda clean -a
conda update -n base -c defaults conda

常用stable-diffusion:
model_key = "stabilityai/stable-diffusion-2-1-base"
model_key = "stabilityai/stable-diffusion-2-base"
model_key = "runwayml/stable-diffusion-v1-5"

训练：
Please set `NCCL_P2P_DISABLE="1"` and `NCCL_IB_DISABLE="1" or use `accelerate launch` which will do this automatically.
export NCCL_P2P_DISABLE="1"
export NCCL_IB_DISABLE="1"
每次重启都要export一下

GPU id:
export CUDA_VISIBLE_DEVICES=1
torch.cuda.set_device(1) 

pip install xformers==0.0.22.post7  # (cu118+torch210)