#!/usr/bin/env bash
set -euo pipefail

SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
source "${SCRIPT_DIR}/../00_setup/env.sh"

: "${DEEPSTARR_DIR:?Set DEEPSTARR_DIR to the local DeepSTARR parquet directory.}"
: "${DIFFUSION_MODEL:?Set DIFFUSION_MODEL to the trained diffusion saved model directory.}"
: "${GENERANNO_BASE_MODEL:?Set GENERANNO_BASE_MODEL to the local GENERanno model.}"
: "${PREDICTOR_MODEL:?Set PREDICTOR_MODEL to the trained predictor best_model directory.}"

mkdir -p "${RESULT_ROOT}/refinement"

cd "${GENERANNO_DIR}"

python3 src/tasks/downstream/discrete_diffusion_refine_evaluate.py \
  --diffusion_model "${DIFFUSION_MODEL}" \
  --base_model_for_code "${GENERANNO_BASE_MODEL}" \
  --predictor_model "${PREDICTOR_MODEL}" \
  --dataset_dir "${DEEPSTARR_DIR}" \
  --output_dir "${RESULT_ROOT}/refinement/paired_refinement_${REFINEMENT_SPLIT:-test}" \
  --split "${REFINEMENT_SPLIT:-test}" \
  --conditioned \
  --target_bucket "${REFINEMENT_TARGET_BUCKET:-high}" \
  --mask_ratios ${REFINEMENT_MASK_RATIOS:-0.01 0.03 0.05 0.10 0.20} \
  --strategies ${REFINEMENT_STRATEGIES:-random entropy logit_gap} \
  --num_sequences "${REFINEMENT_NUM_SEQUENCES:-1024}" \
  --num_samples_per_sequence "${REFINEMENT_NUM_SAMPLES:-8}" \
  --sequence_length "${SEQUENCE_LENGTH:-246}" \
  --num_diffusion_steps "${REFINEMENT_DIFFUSION_STEPS:-32}" \
  --batch_size "${REFINEMENT_BATCH_SIZE:-16}" \
  --predictor_batch_size "${PREDICTOR_SCORE_BATCH_SIZE:-256}" \
  --max_length "${DIFFUSION_MAX_LENGTH:-256}" \
  --temperature "${DIFFUSION_TEMPERATURE:-1.0}" \
  --bf16 \
  --attn_implementation "${ATTN_IMPLEMENTATION:-sdpa}"

echo "Paired refinement evaluation completed: ${RESULT_ROOT}/refinement/paired_refinement_${REFINEMENT_SPLIT:-test}"