| import os |
| import matplotlib.pyplot as plt |
|
|
| exp_dir = "../exp_generalize" |
| selected_epoch = 1 |
|
|
| pretrain_ratios = [] |
| seen_neural2symbolic_accuracy = [] |
| seen_symbolic2neural_accuracy = [] |
| unseen_neural2symbolic_accuracy = [] |
| unseen_symbolic2neural_accuracy = [] |
|
|
| for exp_name in os.listdir(exp_dir): |
| exp_path = os.path.join(exp_dir, exp_name) |
| train_ratio = exp_name.split("pretrain-")[1] |
| find_seen_neural2symbolic_accuracy = False |
| find_unseen_neural2symbolic_accuracy = False |
| find_seen_symbolic2neural_accuracy = False |
| find_unseen_symbolic2neural_accuracy = False |
| if os.path.isdir(exp_path): |
| neural2symbolic_path = os.path.join(exp_path, f"epoch{selected_epoch}", "neural2symbolic.log") |
| symbolic2neural_path = os.path.join(exp_path, f"epoch{selected_epoch}", "symbolic2neural.log") |
| try: |
| with open(neural2symbolic_path, "r") as f: |
| |
| lines = f.readlines() |
| for line in lines: |
| if "accuracy on seen task" in line: |
| find_seen_neural2symbolic_accuracy = True |
| this_seen_neural2symbolic_accuracy = float(line.split("samples: ")[1].split(" ")[0]) |
| elif "accuracy on unseen task" in line: |
| find_unseen_neural2symbolic_accuracy = True |
| this_unseen_neural2symbolic_accuracy = float(line.split("samples: ")[1].split(" ")[0]) |
| with open(symbolic2neural_path, "r") as f: |
| lines = f.readlines() |
| for line in lines: |
| if "accuracy on seen task" in line: |
| find_seen_symbolic2neural_accuracy = True |
| this_seen_symbolic2neural_accuracy = float(line.split("samples: ")[1].split(" ")[0]) |
| elif "accuracy on unseen task" in line: |
| find_unseen_symbolic2neural_accuracy = True |
| this_unseen_symbolic2neural_accuracy = float(line.split("samples: ")[1].split(" ")[0]) |
| except Exception as e: |
| print(f"Error reading file {neural2symbolic_path} or {symbolic2neural_path}: {e}") |
| continue |
| |
| if find_seen_neural2symbolic_accuracy and find_unseen_neural2symbolic_accuracy and find_seen_symbolic2neural_accuracy and find_unseen_symbolic2neural_accuracy: |
| pretrain_ratios.append(float(train_ratio)) |
| seen_neural2symbolic_accuracy.append(this_seen_neural2symbolic_accuracy) |
| unseen_neural2symbolic_accuracy.append(this_unseen_neural2symbolic_accuracy) |
| seen_symbolic2neural_accuracy.append(this_seen_symbolic2neural_accuracy) |
| unseen_symbolic2neural_accuracy.append(this_unseen_symbolic2neural_accuracy) |
|
|
| |
| pretrain_ratios, seen_neural2symbolic_accuracy, unseen_neural2symbolic_accuracy, seen_symbolic2neural_accuracy, unseen_symbolic2neural_accuracy \ |
| = zip(*sorted(zip(pretrain_ratios, seen_neural2symbolic_accuracy, unseen_neural2symbolic_accuracy, seen_symbolic2neural_accuracy, unseen_symbolic2neural_accuracy), key=lambda x: x[0])) |
|
|
| plt.plot(pretrain_ratios, seen_neural2symbolic_accuracy, label="seen neural2symbolic", color='#2ecc71') |
| plt.plot(pretrain_ratios, unseen_neural2symbolic_accuracy, label="unseen neural2symbolic", color='#3498db') |
| plt.plot(pretrain_ratios, seen_symbolic2neural_accuracy, label="seen symbolic2neural", color='#e74c3c') |
| plt.plot(pretrain_ratios, unseen_symbolic2neural_accuracy, label="unseen symbolic2neural", color='#f39c12') |
|
|
| plt.legend(loc='upper right') |
| plt.savefig("accuracy.pdf") |
|
|
|
|
