Spaces:
Sleeping
Sleeping
File size: 3,315 Bytes
f51ee1f | 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 | #!/usr/bin/env python3
"""Inspect and summarize MP statistics data file."""
import argparse
import numpy as np
def main():
parser = argparse.ArgumentParser(description="Inspect MP statistics data")
parser.add_argument("datafile", help="Path to mp_statistics.npz file")
parser.add_argument("--model", help="Show detailed stats for a specific model")
parser.add_argument("--compare", nargs="+", help="Compare predictions across models")
args = parser.parse_args()
data = np.load(args.datafile, allow_pickle=True)
model_names = [k for k in data.files]
print(f"MP Statistics Data File: {args.datafile}")
print(f"Models: {len(model_names)}")
print()
if args.model:
if args.model not in model_names:
print(f"Model {args.model} not found. Available: {', '.join(model_names)}")
return
model = data[args.model].item()
print(f"Model: {args.model}")
print(f" d_model: {model['d_model']}")
print(f" d_head: {model['d_head']}")
print(f" n_heads: {model['n_heads']}")
print(f" n_layers: {model['n_layers']}")
print(f" gamma: {model['d_head'] / model['d_model']:.4f}")
print()
print("MC Statistics (mean ± std):")
for stat_name, values in model['stats'].items():
mean = np.mean(values)
std = np.std(values)
print(f" {stat_name:>35s}: {mean:10.3f} ± {std:8.3f}")
print()
print("MP Predictions:")
for k, v in model['mp_predictions'].items():
print(f" {k:>35s}: {v:10.3f}")
print()
print("PME (Product Multiplicative Ensemble) Predictions:")
for k, v in model['pme_predictions'].items():
print(f" {k:>35s}: {v:10.3f}")
elif args.compare:
models_to_compare = [m for m in args.compare if m in model_names]
if not models_to_compare:
print(f"None of the requested models found. Available: {', '.join(model_names)}")
return
print(f"{'Model':<20s} {'d_model':>8s} {'d_head':>7s} {'gamma':>7s} " +
f"{'PR (MC)':>10s} {'PR (MP)':>10s} {'PR (PME)':>10s}")
print("-" * 93)
for model_name in models_to_compare:
model = data[model_name].item()
pr_mc = np.mean(model['stats']['participation_ratio'])
pr_mp = model['mp_predictions']['participation_ratio']
pr_pme = model['pme_predictions']['participation_ratio']
gamma = model['d_head'] / model['d_model']
print(f"{model_name:<20s} {model['d_model']:8d} {model['d_head']:7d} {gamma:7.4f} " +
f"{pr_mc:10.3f} {pr_mp:10.3f} {pr_pme:10.3f}")
else:
print(f"{'Model':<20s} {'d_model':>8s} {'d_head':>7s} {'n_heads':>8s} " +
f"{'n_layers':>9s} {'gamma':>7s} {'n_samples':>10s}")
print("-" * 90)
for model_name in sorted(model_names):
model = data[model_name].item()
n_samples = len(model['stats']['max'])
gamma = model['d_head'] / model['d_model']
print(f"{model_name:<20s} {model['d_model']:8d} {model['d_head']:7d} " +
f"{model['n_heads']:8d} {model['n_layers']:9d} {gamma:7.4f} {n_samples:10d}")
if __name__ == "__main__":
main()
|