scripts/train_cta.sh

#!/usr/bin/env bash
# Train CTA on 8 GPUs (or whatever CUDA_VISIBLE_DEVICES limits us to).
set -euo pipefail
cd "$(dirname "$0")/.."

# load .env if present
[ -f .env ] && set -a && . ./.env && set +a

python3 src/train.py \
    method=cta \
    data=fairtalking \
    trainer=ddp \
    backbone=timesformer \
    "$@"