| #!/bin/bash |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
|
|
| set -e |
| cd "$(dirname "$0")/.." |
|
|
| DATA_DIR=${DATA_DIR:-./data} |
| CONFIG=${CONFIG:-configs/training_config.yaml} |
| SPLIT=${SPLIT:-train} |
| MAX_PER_SITE=${MAX_PER_SITE:-500} |
|
|
| echo "============================================================" |
| echo " Trigger-Off: Data Preparation" |
| echo "============================================================" |
| echo " DATA_DIR = $DATA_DIR" |
| echo " CONFIG = $CONFIG" |
| echo " SPLIT = $SPLIT" |
| echo " MAX_PER_SITE = $MAX_PER_SITE" |
| echo "------------------------------------------------------------" |
|
|
| |
| echo "" |
| echo "[1/3] Downloading osunlp/Multimodal-Mind2Web (split=$SPLIT) …" |
| python -m src.data_pipeline.mind2web_loader \ |
| --output_dir "$DATA_DIR/mind2web_raw" \ |
| --split "$SPLIT" \ |
| --save_images |
|
|
| |
| echo "" |
| echo "[2/3] Filtering and balancing by site …" |
| python -m src.data_pipeline.site_filter \ |
| --input_dir "$DATA_DIR/mind2web_raw" \ |
| --output_dir "$DATA_DIR/processed" \ |
| --max_per_site "$MAX_PER_SITE" |
|
|
| |
| echo "" |
| echo "[3/3] Building Trigger-Off poisoned dataset …" |
| python -m src.data_pipeline.dataset_builder \ |
| --input_dir "$DATA_DIR/processed" \ |
| --output_dir "$DATA_DIR/poisoned" \ |
| --config "$CONFIG" |
|
|
| echo "" |
| echo "============================================================" |
| echo " Data preparation complete." |
| echo " Train JSON : $DATA_DIR/poisoned/train.json" |
| echo " Test JSON : $DATA_DIR/poisoned/test.json" |
| echo "============================================================" |
|
|