#!/bin/bash
#PJM -L rscgrp=b-batch
#PJM -L gpu=1
#PJM -L elapse=24:00:00
#PJM -N ccfm_v2_cached
#PJM -j
#PJM -o logs/ccfm_v2_cached_%j.out

module load cuda/12.2.2
module load cudnn/8.9.7
module load gcc-toolset/12

source /home/hp250092/ku50001222/qian/aivc/lfj/stack_env/bin/activate

cd /home/hp250092/ku50001222/qian/aivc/lfj/transfer/code/CCFM

export PYTORCH_CUDA_ALLOC_CONF=max_split_size_mb:256

echo "=========================================="
echo "Job ID:    $PJM_JOBID"
echo "Job Name:  $PJM_JOBNAME"
echo "Start:     $(date)"
echo "Node:      $(hostname)"
echo "GPU:       $(nvidia-smi --query-gpu=name,memory.total --format=csv,noheader 2>/dev/null || echo 'N/A')"
echo "Run:       CCFM v2 cached (KV fix + loss fix + EMA + RK4 + logit-normal + warmup + scGPT cache)"
echo "=========================================="

accelerate launch --num_processes=1 scripts/run_cascaded.py \
    --data-name norman \
    --d-model 128 \
    --nhead 8 \
    --nlayers 4 \
    --batch-size 48 \
    --lr 5e-5 \
    --steps 200000 \
    --fusion-method differential_perceiver \
    --perturbation-function crisper \
    --noise-type Gaussian \
    --infer-top-gene 1000 \
    --n-top-genes 5000 \
    --use-mmd-loss \
    --gamma 0.5 \
    --split-method additive \
    --fold 1 \
    --scgpt-dim 512 \
    --bottleneck-dim 128 \
    --latent-weight 1.0 \
    --choose-latent-p 0.4 \
    --dh-depth 2 \
    --print-every 10000 \
    --topk 30 \
    --use-negative-edge \
    --ema-decay 0.9999 \
    --t-sample-mode logit_normal \
    --t-expr-mean 0.0 \
    --t-expr-std 1.0 \
    --t-latent-mean 0.0 \
    --t-latent-std 1.0 \
    --warmup-steps 2000 \
    --ode-method rk4 \
    --scgpt-cache-path scgpt_cache_norman.h5 \
    --result-path ./result

echo "=========================================="
echo "Finished:  $(date)"
echo "=========================================="