From 466e1256efaae032ea25a7a9dfd8db5e35832ea1 Mon Sep 17 00:00:00 2001
From: Chen <joshua86@cc-login2.campuscluster.illinois.edu>
Date: Sun, 15 Mar 2026 01:08:37 -0500
Subject: [PATCH] cluster bash script, yaml for conda env, docs

---
 clustersetup.md          | 41 ++++++++++++++++++++++++++
 pyhealth_cluster_env.yml | 63 ++++++++++++++++++++++++++++++++++++++++
 setup.sh                 | 53 +++++++++++++++++++++++++++++++++
 3 files changed, 157 insertions(+)
 create mode 100644 clustersetup.md
 create mode 100644 pyhealth_cluster_env.yml
 create mode 100644 setup.sh

diff --git a/clustersetup.md b/clustersetup.md
new file mode 100644
index 000000000..93c3af445
--- /dev/null
+++ b/clustersetup.md
@@ -0,0 +1,41 @@
+# Running PyHealth on the Campus Cluster
+
+## Quick Start
+
+```bash
+git clone https://github.com/Multimodal-PyHealth/PyHealth.git
+chmod +x setup.sh
+./setup.sh
+```
+
+## Data Paths
+
+All MIMIC-4 data is under `/projects/illinois/eng/cs/jimeng/physionet.org/files/`:
+
+| Data | Path |
+| EHR | `/projects/illinois/eng/cs/jimeng/physionet.org/files/mimiciv/2.2` |
+| Clinical Notes | `/projects/illinois/eng/cs/jimeng/physionet.org/files/mimic-note` |
+| Chest X-rays | `/projects/illinois/eng/cs/jimeng/physionet.org/files/mimic-cxr-jpg/2.1.0` |
+
+**Important:** `NOTE_ROOT` should be `.../mimic-note` (not `.../mimic-note/note`). The config YAML appends `note/` automatically.
+
+Set `CACHE_DIR` to your own writable directory: `/u/<NetID>/pyhealth_cache`
+
+## Running on a Compute Node
+
+Run on a compute node:
+Slurm command example:
+```bash
+srun --account=jimeng-cs-eng --partition=eng-research-gpu --time=00:10:00 --gres=gpu:1 --pty bash
+```
+
+Once on the compute node, re-activate and run:
+
+```bash
+module load miniconda3/24.9.2
+conda activate pyhealth2
+cd ~/PyHealth
+python examples/mortality_prediction/multimodal_mimic4.py
+```
+
+
diff --git a/pyhealth_cluster_env.yml b/pyhealth_cluster_env.yml
new file mode 100644
index 000000000..5ea5d60a1
--- /dev/null
+++ b/pyhealth_cluster_env.yml
@@ -0,0 +1,63 @@
+name: pyhealth2
+channels:
+  - pytorch
+  - nvidia
+  - conda-forge
+  - defaults
+
+dependencies:
+  # ── Core Python (PyHealth 2.0 requires >=3.12, <3.14) ──
+  - python=3.12
+
+  # ── PyTorch + CUDA ──────────────────────────────────────
+  # Run `nvidia-smi` on the cluster to check your CUDA version.
+  # Cluster has CUDA driver 13.0, so cu126 runtime is compatible.
+  # Pinning both to same minor version avoids mismatch errors.
+  - pytorch=2.7.1
+  - torchvision=0.22.1
+  - pytorch-cuda=12.6
+
+  # ── Scientific stack ────────────────────────────────────
+  - numpy>=2.0
+  - scipy
+  - scikit-learn
+  - pandas
+  - pyarrow>=15.0
+  - dask>=2024.1.0
+
+  # ── Build / utility ─────────────────────────────────────
+  - pip
+  - tqdm
+  - networkx
+  - packaging
+  - pydantic>=2.0
+  - urllib3>=2.0
+
+  # ── Jupyter (for notebook dev on cluster) ───────────────
+  - jupyterlab
+  - ipykernel
+  - ipywidgets
+
+  # ── pip-only packages ───────────────────────────────────
+  - pip:
+      # PyHealth: editable install from the forked repo clone
+      # Run this AFTER creating the env:
+      #   conda activate pyhealth2
+      #   cd ~/PyHealth
+      #   pip install -e .
+      #
+      # Dependencies that pyproject.toml pulls in:
+      - polars>=1.30
+      - litdata>=0.2
+      - narwhals>=1.0
+      - more-itertools>=10.0
+      - einops>=0.8.0
+      - linear-attention-transformer>=0.19.1
+      - pandarallel>=1.6
+      - mne>=1.8
+      - transformers>=4.40
+      - accelerate
+      - peft
+      - rdkit
+      - ogb>=1.3.5
+      - xgboost
\ No newline at end of file
diff --git a/setup.sh b/setup.sh
new file mode 100644
index 000000000..18b59eafa
--- /dev/null
+++ b/setup.sh
@@ -0,0 +1,53 @@
+#  PyHealth Cluster Environment Setup
+#  Usage: ./setup.sh
+
+set -e
+
+ENV_NAME="pyhealth2"
+ENV_FILE="pyhealth_cluster_env.yml"
+
+GREEN='\033[0;32m'
+YELLOW='\033[1;33m'
+NC='\033[0m'
+
+echo -e "${GREEN}  PyHealth Cluster Setup${NC}"
+
+#  Load miniconda 
+if ! command -v conda &> /dev/null; then
+    echo -e "${YELLOW}Loading miniconda module...${NC}"
+    module load miniconda3/24.9.2
+    source ~/.bashrc
+fi
+
+#  Create or skip conda env 
+if conda info --envs 2>/dev/null | grep -q "${ENV_NAME}"; then
+    echo -e "${GREEN}Conda environment '${ENV_NAME}' already exists. Skipping creation.${NC}"
+else
+    echo -e "${YELLOW}Creating conda environment '${ENV_NAME}'...${NC}"
+    if [ ! -f "$ENV_FILE" ]; then
+        echo "Error: ${ENV_FILE} not found. Make sure you're in the PyHealth repo root."
+        exit 1
+    fi
+    conda env create -f ${ENV_FILE}
+fi
+
+#  Activate env 
+eval "$(conda shell.bash hook)"
+conda activate ${ENV_NAME}
+echo -e "${GREEN}Activated ${ENV_NAME} (Python: $(python --version))${NC}"
+
+#  Install PyHealth editable if needed 
+if python -c "import pyhealth" 2>/dev/null; then
+    echo -e "${GREEN}PyHealth already installed. Skipping.${NC}"
+else
+    echo -e "${YELLOW}Installing PyHealth in editable mode...${NC}"
+    pip install -e .
+fi
+
+#  Verify 
+echo ""
+echo -e "${GREEN}Verification:${NC}"
+python -c "import pyhealth; print(f'  PyHealth: OK')"
+python -c "import torch; print(f'  PyTorch {torch.__version__}, CUDA available: {torch.cuda.is_available()}')"
+echo ""
+echo -e "${GREEN}Setup complete!${NC}"
\ No newline at end of file