1. Transfer Your Local Repo to TACC Scratch On your local terminal (Mac):

scp -r /Users/romirpatel/ct-diffusionmodelbench romirpatel@vista.tacc.utexas.edu:/scratch/10936/romirpatel/

2. SSH to Vista

ssh romirpatel@vista.tacc.utexas.edu

3. Free Up Home Directory Quota (if needed) Check quota:

myquota

If home quota is full, clean cache:

rm -rf ~/.cache/huggingface ~/.cache/torch

Move big files from home to scratch as needed.

4. Launch Interactive GPU Session

idev -p gh-dev -N 1 -n 1 -t 02:00:00

5. Load Required Modules Once inside the compute node session:

module load gcc
module load cuda
module load python3

6. Navigate to Your Project on Scratch

cd /scratch/10936/romirpatel/ct-diffusionmodelbench

7. Setup and Activate Virtual Environment If you haven't created the virtual environment:

python3 -m venv venv
source venv/bin/activate
pip install --upgrade pip
pip install -r requirements.txt

If you already created one:

source venv/bin/activate

8. Set Hugging Face Cache Directory Set these environment variables to avoid filling home directory:

export HF_HOME=$SCRATCH/huggingface_cache
export HF_HUB_CACHE=$HF_HOME
export TRANSFORMERS_CACHE=$HF_HOME
mkdir -p $HF_HOME

9. In order to run inference, run the following command:

python Inference/chat_finetuned.py