Source

This repository is cloned and adapted from PartField, with modifications for processing the PartSlip dataset for student model learning.

Usage on HPC

Environment

module load conda

conda env create -f environment.yml

conda activate partfield

pip install huggingface_hub

Download pretrained model

mkdir model

cd model

hf download mikaelaangel/partfield-ckpt Make sure the model is in /model

Submit and track a job

In the terminal, use the following command, where BATCH_ID should be specified.

sh pipeline_infere_auto.sh BATCH_ID

The pipeline include downloading a batch of data and infere. The data directory and output directory are listed in pipeline_infere.sh. Please change them accordingly. SCRATCH_BASE="/scratch/eecs442f25_class_root/eecs442f25_class/jonzhe" -> SCRATCH_BASE="/scratch/eecs442f25_class_root/eecs442f25_class/YOUR_USER_NAME"

You can see the job running status by squeue -u $USER.

The output log will be partfield_pipeline_${BATCH_ID}.log in the current directory.