File size: 6,286 Bytes
21d04bc | 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 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 | """
Launch separator-attention and random-target intervention experiments across 8 GPUs.
Each GPU runs N trials, collecting per-number success data.
Assembles two plots:
1. Success probability per number when intervening at separator-attending positions
2. Success probability per number when intervening with random target
"""
import os
import subprocess
import sys
import time
import numpy as np
import matplotlib
matplotlib.use('Agg')
import matplotlib.pyplot as plt
SCRIPT_DIR = os.path.dirname(os.path.abspath(__file__))
OUTPUT_BASE = os.path.join(SCRIPT_DIR, 'outputs')
PLOT_DIR = os.path.join(OUTPUT_BASE,
'plots_V256_B16_LR3e-2_MI200000_E64_H1_L2_ds1337_is1337_ckpt200000')
TMP_DIR = os.path.join(OUTPUT_BASE, 'tmp_results', 'separator_random')
CKPT = os.path.join(SCRIPT_DIR,
'sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__final.pt')
NUM_GPUS = 8
TRIALS_PER_GPU = 1000
INTENSITIES = [2.0, 6.0, 10.0]
TAG = 'V=256 B=16 lr=0.03 iters=200000 dseed=1337 iseed=1337'
def launch_workers():
os.makedirs(TMP_DIR, exist_ok=True)
log_dir = os.path.join(OUTPUT_BASE, 'separator_logs')
os.makedirs(log_dir, exist_ok=True)
procs = []
for g in range(NUM_GPUS):
out = os.path.join(TMP_DIR, f'gpu{g}.npz')
if os.path.exists(out):
continue
lf = open(os.path.join(log_dir, f'gpu{g}.log'), 'w')
proc = subprocess.Popen(
[sys.executable, os.path.join(SCRIPT_DIR, 'separator_intervention_worker.py'),
'--ckpt', CKPT, '--gpu', str(g),
'--trials', str(TRIALS_PER_GPU), '--out', out],
stdout=lf, stderr=subprocess.STDOUT, cwd=SCRIPT_DIR)
procs.append((proc, lf, g))
return procs
def wait_for_workers(procs):
t0 = time.time()
while any(p.poll() is None for p, _, _ in procs):
time.sleep(10)
done = sum(1 for g in range(NUM_GPUS)
if os.path.exists(os.path.join(TMP_DIR, f'gpu{g}.npz')))
print(f" [{time.time()-t0:.0f}s] {done}/{NUM_GPUS} GPUs done", flush=True)
for proc, lf, g in procs:
lf.close()
if proc.returncode != 0:
print(f" WARN: GPU {g} exited with code {proc.returncode}", flush=True)
print(f"All workers done in {time.time()-t0:.0f}s", flush=True)
def load_and_combine():
all_sep, all_rand = [], []
for g in range(NUM_GPUS):
f = os.path.join(TMP_DIR, f'gpu{g}.npz')
if not os.path.exists(f):
continue
d = np.load(f)
if len(d['sep_data']) > 0:
all_sep.append(d['sep_data'])
if len(d['rand_data']) > 0:
all_rand.append(d['rand_data'])
sep = np.concatenate(all_sep) if all_sep else np.empty((0, 3), dtype=np.int32)
rand = np.concatenate(all_rand) if all_rand else np.empty((0, 3), dtype=np.int32)
print(f"Combined: sep={len(sep)} records, rand={len(rand)} records")
return sep, rand
def plot_per_number(data, title_prefix, filename, tag):
"""Plot success probability per number for each intensity."""
os.makedirs(PLOT_DIR, exist_ok=True)
colors = {2.0: '#1f77b4', 6.0: '#ff7f0e', 10.0: '#d62728'}
fig, axes = plt.subplots(2, 1, figsize=(14, 8),
gridspec_kw={'height_ratios': [3, 1]})
ax = axes[0]
for intens in INTENSITIES:
mask = data[:, 1] == intens
subset = data[mask]
if len(subset) == 0:
continue
xs, ys = [], []
for n_val in range(256):
nm = subset[:, 0] == n_val
count = nm.sum()
if count >= 10:
xs.append(n_val)
ys.append(subset[nm, 2].mean())
ax.plot(xs, ys, color=colors.get(intens, '#333'),
linewidth=0.8, alpha=0.6, label=f'raw int={intens}')
if len(xs) >= 11:
raw_arr = np.full(256, np.nan)
for x, y in zip(xs, ys):
raw_arr[x] = y
win = 11
padded = np.nan_to_num(raw_arr, nan=0.5)
smoothed = np.convolve(padded, np.ones(win) / win, mode='same')
valid = ~np.isnan(raw_arr)
ax.plot(np.arange(256)[valid], smoothed[valid],
color=colors.get(intens, '#333'), linewidth=2.5,
linestyle='--', label=f'smoothed int={intens}')
ax.set_ylabel('Success Probability', fontsize=12)
ax.set_title(f'{title_prefix} (Layer 0)\n{tag}', fontsize=12, fontweight='bold')
ax.legend(fontsize=8, ncol=2, loc='lower left')
ax.grid(True, alpha=0.3)
ax.set_ylim(-0.05, 1.1)
ax.set_xlim(0, 255)
ax2 = axes[1]
max_intens = max(INTENSITIES)
mask_hi = data[:, 1] == max_intens
counts = np.array([(mask_hi & (data[:, 0] == n)).sum() for n in range(256)])
ax2.bar(range(256), counts, width=1, color='#666', alpha=0.5)
ax2.set_xlabel('Number in Vocabulary', fontsize=12)
ax2.set_ylabel('Sample Count', fontsize=10)
ax2.set_xlim(0, 255)
ax2.grid(True, alpha=0.2)
fig.tight_layout()
out_path = os.path.join(PLOT_DIR, filename)
fig.savefig(out_path, dpi=200, bbox_inches='tight')
plt.close()
print(f"Plot saved: {out_path}")
def main():
t0 = time.time()
print("=" * 60)
print("SEPARATOR & RANDOM INTERVENTION EXPERIMENT (Layer 0)")
print("=" * 60)
cached = sum(1 for g in range(NUM_GPUS)
if os.path.exists(os.path.join(TMP_DIR, f'gpu{g}.npz')))
print(f"Cached GPUs: {cached}/{NUM_GPUS}")
procs = launch_workers()
if procs:
print(f"Launched {len(procs)} workers")
wait_for_workers(procs)
else:
print("All workers already cached")
sep, rand = load_and_combine()
if len(sep) > 0:
plot_per_number(sep,
'Intervention Success when Attending to Separator',
'intervention_pernumber_separator_layer0.png', TAG)
else:
print("WARNING: No separator-attending data collected!")
if len(rand) > 0:
plot_per_number(rand,
'Intervention Success with Random Target',
'intervention_pernumber_random_layer0.png', TAG)
elapsed = time.time() - t0
print(f"\nTotal time: {elapsed:.0f}s ({elapsed/60:.1f}m)")
if __name__ == '__main__':
main()
|