DeepTCM/run_cl.sh at main · spartanjoax/DeepTCM · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
#!/usr/bin/env bash
# run_cl_experiments.sh — Full Ch06 CL experiment suite
#
# Runs:
#   1. train_separate.py  – per-condition separate-model baseline (all datasets, all signal groups)
#   2. train_cl.py        – CL strategies (all scenarios, all strategies, 10 seeds)
#
# Usage (from repo root):
#   bash run_cl.sh                   # full run, 10 seeds
#   bash run_cl.sh --smoke-test      # 1-epoch sanity check, seed 23452 only
#
# Results land in:
#   continual/results/separate/            (train_separate.py)
#   continual/results/seed_<N>/            (train_cl.py)
#
# Seed selection policy (for reviewer transparency):
#   10 seeds drawn via: random.seed(2026); random.sample(range(10000, 99999), 10)
#   Sorted: 23452 25613 39266 51875 75865 77085 82949 88778 91459 94829
#   Reproducible and non-cherry-picked. 95% CI uses t_{9, 0.025} = 2.262.
#
# Requires: conda environment "torch"
# ---------------------------------------------------------------------------

set -euo pipefail

SMOKE_TEST=false
if [[ "${1:-}" == "--smoke-test" ]]; then
    SMOKE_TEST=true
fi

# 10 random seeds: random.seed(2026); random.sample(range(10000, 99999), 10)
SEEDS=(23452 25613 39266 51875 75865 77085 82949 88778 91459 94829)

# ---------------------------------------------------------------------------
# 1. Separate-model LOCO-CV baseline
# ---------------------------------------------------------------------------
echo "========================================================"
echo "  STEP 1 — Separate-model LOCO-CV baseline"
echo "========================================================"

if $SMOKE_TEST; then
    python train_separate.py \
        --datasets all \
        --signal_groups AC DC AC_table DC_table internals internals_v2 internals_v3 \
        --seed "${SEEDS[0]}" \
        --smoke_test
else
    for SEED in "${SEEDS[@]}"; do
        echo ""
        echo "  --- Seed ${SEED} ---"
        python train_separate.py \
            --datasets all \
            --signal_groups AC DC AC_table DC_table internals internals_v2 internals_v3 \
            --seed "$SEED"
        echo "  [DONE] seed ${SEED}"
    done
fi

echo "  [DONE] train_separate.py"
echo ""

# ---------------------------------------------------------------------------
# 2. CL strategies — one run per seed
# ---------------------------------------------------------------------------
echo "========================================================"
echo "  STEP 2 — Continual Learning strategies"
echo "========================================================"

if $SMOKE_TEST; then
    python train_cl.py \
        --scenarios scenario_1 scenario_2 scenario_3 scenario_4 scenario_5 scenario_6 \
        --seed "${SEEDS[0]}" \
        --smoke-test
else
    for SEED in "${SEEDS[@]}"; do
        echo ""
        echo "  --- Seed ${SEED} ---"
        python train_cl.py \
            --scenarios scenario_1 scenario_2 scenario_3 scenario_4 scenario_5 scenario_6 \
            --seed "$SEED"
        echo "  [DONE] seed ${SEED}"
    done
fi

echo ""
echo "========================================================"
echo "  All CL experiments complete."
echo "  Results: continual/results/"
echo "========================================================"