File size: 3,771 Bytes
3e10edb |
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 |
#!/usr/bin/env python
# -*- coding: utf-8 -*-
import argparse
import os
import equinox as eqx
import jax
import jax.numpy as jnp
import numpy as np
import pandas as pd
from popgym_arcade.baselines.model.builder import QNetworkRNN
from popgym_arcade.baselines.utils import get_terminal_saliency_maps
def run_multiple_seeds_and_save_csv(config, seeds, max_steps=200, output_csv=None):
"""
Run saliency analysis on multiple seeds and save the results in a CSV file.
Args:
config: Configuration dictionary
seeds: List of seeds to run
max_steps: Maximum number of steps for each episode
output_csv: Path to save the CSV file (default: auto-generated based on config)
Returns:
Path to the saved CSV file
"""
# Create a default output path if none provided
if output_csv is None:
output_csv = f'saliency_results_{config["MEMORY_TYPE"]}_{config["ENV_NAME"]}_Partial={config["PARTIAL"]}.csv'
# List to store results
all_results = []
# Store saliency distributions for each seed
for seed_value in seeds:
print(f"Processing seed {seed_value}...")
# Update config with current seed
config["SEED"] = seed_value
# Create the model path for this seed
model_path = f"pkls_gradients/PQN_RNN_{config['MEMORY_TYPE']}_{config['ENV_NAME']}_model_Partial={config['PARTIAL']}_SEED={config['MODEL_SEED']}.pkl"
# Initialize random key for this seed
rng = jax.random.PRNGKey(seed_value)
# Initialize and load the model
network = QNetworkRNN(
rng, rnn_type=config["MEMORY_TYPE"], obs_size=config["OBS_SIZE"]
)
# try:
model = eqx.tree_deserialise_leaves(model_path, network)
# Define path for saving the distribution for this seed
dist_save_path = f'dist_{config["MEMORY_TYPE"]}_{config["ENV_NAME"]}_Partial={config["PARTIAL"]}_SEED={seed_value}.npy'
# Run terminal saliency analysis
grads_obs = get_terminal_saliency_maps(
rng,
model,
config,
)
# print(grads_obs.shape)
# grads_obs = grads_obs.squeeze(1)
grads_obs = jnp.abs(grads_obs).sum(axis=(1, 2, 3))
dist = grads_obs / grads_obs.sum()
print(dist.sum())
# Convert JAX array to numpy for DataFrame
dist_np = np.array(dist)
# Create result dictionary
result = {
"seed": seed_value,
"distribution": dist_np,
"length": len(dist_np),
"dist_path": dist_save_path,
}
all_results.append(result)
print(f"Seed {seed_value} completed. Distribution length: {len(dist_np)}")
# except Exception as e:
# raise e
# # print(f"Error processing seed {seed_value}: {e}")
# Process results for CSV format
csv_data = []
max_length = max([r["length"] for r in all_results]) if all_results else 0
for result in all_results:
# Pad distribution to max length if needed
padded_dist = np.zeros(max_length)
padded_dist[: result["length"]] = result["distribution"]
# Create row data
row = {
"seed": result["seed"],
"length": result["length"],
"dist_path": result["dist_path"],
}
# Add each position value as a separate column
for i in range(max_length):
norm_pos = i / max_length if max_length > 0 else 0
row[f"pos_{norm_pos:.3f}"] = padded_dist[i]
csv_data.append(row)
# Create DataFrame and save to CSV
df = pd.DataFrame(csv_data)
df.to_csv(output_csv, index=False)
print(f"Results saved to {output_csv}")
return output_csv
|