We recommend using Linux to run the code, as testing has shown that the training time required is shorter compared to Windows.
Using Anaconda to create environment.
conda create -n pl312 python=3.12 conda activate pl312
Install packages. Notice: PyTorch use the CPU version.
pip install pennylane --upgrade pip install torch torchvision torchaudio scikit-learn opencv-python matplotlib tqdm torchsummary conda install -c conda-forge jupyterlab
If you’re looking to train on all categories, we recommend dividing the code into several parts and running the training process (tasks) for different superclasses. Each process requires relatively few computational resources, so running multiple processes simultaneously can improve efficiency. This requires you to manually modify some parts of the code.
For CPUs with good performance, we recommend running up to 10 training tasks simultaneously, with each task handling two superclasses.