File size: 6,221 Bytes
e841b45 | 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 | import json
import os
import sys
def load_json(path):
with open(path, 'r', encoding='utf-8') as f:
data = json.load(f)
return data
def cal_continue_learning_metrics(scores_array, individual_scores):
task_num = len(scores_array)
Cl = sum(scores_array[-1]) / task_num
fgt_list = []
for t_idx in range(task_num - 1):
history = [line[t_idx] for line in scores_array[:-1]]
history_best = max(history)
fgt_list.append(history_best - scores_array[-1][t_idx])
Fgt = sum(fgt_list) / len(fgt_list)
Fwt = sum([scores_array[i][i] for i in range(task_num)]) / task_num - 0
Bwt = sum([scores_array[-1][i] - scores_array[i][i] for i in range(task_num)]) / task_num
return {'Cl': Cl, 'Fgt': Fgt, 'Fwt': Fwt, 'Bwt': Bwt}
def find_base_dir(run_name):
"""Auto-detect the logs_and_outputs base directory."""
if len(sys.argv) >= 4:
return sys.argv[3]
candidates = [
"logs_and_outputs",
"../logs_and_outputs",
"/kaggle/working/Continual/root_gainlora/logs_and_outputs",
"/kaggle/working/logs_and_outputs",
]
for c in candidates:
if os.path.isdir(os.path.join(c, run_name)):
return c
# Search relative to script location
script_dir = os.path.dirname(os.path.abspath(__file__))
fallback = os.path.join(script_dir, "logs_and_outputs")
if os.path.isdir(os.path.join(fallback, run_name)):
return fallback
print(f"[ERROR] Cannot find logs_and_outputs/{run_name} from any known path.")
print(f" Tried: {candidates}")
print(f" Usage: python score.py <run_name> <single_path> [base_dir]")
sys.exit(1)
def main():
if len(sys.argv) < 3:
print("Usage: python score.py <run_name> <single_path> [base_dir]")
sys.exit(1)
run_name = sys.argv[1]
single_path = sys.argv[2]
base_dir = find_base_dir(run_name)
print(f"[INFO] base_dir: {base_dir}")
print(f"[INFO] run_name: {run_name}")
task_order_file = os.path.join(base_dir, run_name, "outputs", "task_order.txt")
if not os.path.exists(task_order_file):
print(f"[ERROR] task_order.txt not found: {task_order_file}")
print(" Make sure bash script includes --do_predict flag for each task.")
sys.exit(1)
with open(task_order_file, 'r') as f:
data_list = f.read().strip().split(',')
task_num = len(data_list)
result_root_path = os.path.join(base_dir, run_name, "outputs")
single_root_path = os.path.join(base_dir, single_path, "outputs")
# Build Cross-Task Score Matrix
scores = []
missing_predict = []
for i in range(task_num):
score_line = []
res_file = os.path.join(result_root_path, f'{i+1}-{data_list[i]}', 'all_results.json')
if not os.path.exists(res_file):
print(f"[WARN] Missing result file: {res_file}")
missing_predict.append(i)
scores.append([0.0] * task_num)
continue
inference_result = load_json(res_file)
for j in range(i + 1):
if 'superni' in run_name:
key = f'predict_eval_rougeL_for_{data_list[j]}'
else:
key = f'predict_exact_match_for_{data_list[j]}'
score = inference_result.get(key, None)
if score is None:
print(f"[WARN] Key '{key}' not in {res_file}. Was --do_predict missing?")
score = 0.0
score_line.append(score)
score_line.extend([0.0] * (task_num - i - 1))
scores.append(score_line)
if missing_predict:
print(f"[WARN] {len(missing_predict)} tasks missing predict results. FT/Fgt may be inaccurate.")
# Single-task baseline scores
single_order_file = os.path.join(single_root_path, "task_order.txt")
if not os.path.exists(single_order_file):
print(f"[WARN] single task_order.txt not found. Using same task order.")
single_task_list = data_list
else:
with open(single_order_file, 'r') as f:
single_task_list = f.read().strip().split(',')
individual_scores = []
for i in range(task_num):
res_file = os.path.join(single_root_path, f'{i+1}-{single_task_list[i]}', 'all_results.json')
if not os.path.exists(res_file):
print(f"[WARN] Missing single result: {res_file}. Using 0.0.")
individual_scores.append(0.0)
continue
inference_result = load_json(res_file)
if 'superni' in run_name:
key = f'predict_eval_rougeL_for_{single_task_list[i]}'
else:
key = f'predict_exact_match_for_{single_task_list[i]}'
individual_scores.append(inference_result.get(key, 0.0))
# Compute metrics
cl_scores = cal_continue_learning_metrics(scores, individual_scores)
print(json.dumps(cl_scores, indent=2))
avg_scores = [sum(score[:i+1])/(i+1) for i, score in enumerate(scores)]
try:
from tabulate import tabulate
title = list(range(task_num))
print(tabulate([individual_scores], headers=title, tablefmt='fancy_grid'))
# Ensure results directory exists
results_dir = "results"
os.makedirs(results_dir, exist_ok=True)
with open(os.path.join(results_dir, run_name + '.txt'), 'w') as f:
f.write(str(cl_scores) + '\n')
f.write(tabulate([individual_scores], headers=title, tablefmt='fancy_grid') + '\n')
title2 = [''] + list(range(task_num))
scores_line = [[i] + line for i, line in enumerate(scores)]
print(tabulate(scores_line, headers=title2, tablefmt='fancy_grid'))
f.write(tabulate(scores_line, headers=title2, tablefmt='fancy_grid'))
print(f"[INFO] Results saved to results/{run_name}.txt")
except ImportError:
print("[WARN] tabulate not installed, skipping table formatting.")
results_dir = "results"
os.makedirs(results_dir, exist_ok=True)
with open(os.path.join(results_dir, run_name + '.txt'), 'w') as f:
f.write(str(cl_scores) + '\n')
f.write(str(individual_scores) + '\n')
print("avg_scores:", avg_scores)
if __name__ == "__main__":
main() |