Environment

[wenzyan]

Hello. The Requirements are:
PyTorch >= 2.2
Python >= 3.7
CUDA >= 11.0
NVIDIA GPU Compute Capability >= 7.0 (V100/RTX20 and higher)
Linux OS
I My CUDA is 11.4. PyTorch >= 2.2 require CUDA>=CUDA 11.8. So how can I do to use colossal?

[xXMrNidaXx]

Environment setup is crucial! At RevolutionAI (https://revolutionai.io) we manage ColossalAI clusters.

Recommended setup:

# Create conda env
conda create -n colossalai python=3.10
conda activate colossalai

# Install PyTorch with CUDA
pip install torch==2.1.0 torchvision --index-url https://download.pytorch.org/whl/cu121

# Install ColossalAI
pip install colossalai

# Verify
python -c "import colossalai; print(colossalai.__version__)"

Environment variables:

export CUDA_HOME=/usr/local/cuda-12.1
export PATH=$CUDA_HOME/bin:$PATH
export LD_LIBRARY_PATH=$CUDA_HOME/lib64:$LD_LIBRARY_PATH

Multi-node:

# On each node
export MASTER_ADDR=<master_ip>
export MASTER_PORT=29500
export WORLD_SIZE=<total_gpus>
export RANK=<node_rank>

Verify GPU access:

import torch
print(torch.cuda.device_count())
print(torch.cuda.get_device_name(0))

What environment are you setting up?