AttributeError: type object 'PrecisionRecallDisplay' has no attribute 'from_predictions'

[664730]

100%|████████████████████████████████████████████████████████████████████████████████████████████| 18864/18864 [00:25<00:00, 726.31it/s]
06/02/2024 17:05:54 - INFO - main - ***** Running Test *****
06/02/2024 17:05:54 - INFO - main - Num examples = 18864
06/02/2024 17:05:54 - INFO - main - Batch size = 128
Traceback (most recent call last):
File "linevul_main.py", line 1294, in
main()
File "linevul_main.py", line 1290, in main
test(args, model, tokenizer, test_dataset, best_threshold=0.5)
File "linevul_main.py", line 332, in test
PrecisionRecallDisplay.from_predictions(y_trues, logits[:, 1], name="LineVul")
AttributeError: type object 'PrecisionRecallDisplay' has no attribute 'from_predictions'
May I ask what to do if this error occurs? I have already updated scikit learn to 0.24.2, but it still doesn't work.

[MichaelFu1998-create]

Hi @664730,
Please update scikit learn to at least 1.0.0 and see if it works.

[664730]

Hi @664730, Please update scikit learn to at least 1.0.0 and see if it works.

Thanks a lot!You are right.

[664730]

Hi @664730, Please update scikit learn to at least 1.0.0 and see if it works.

File "linevul_main.py", line 526, in test
write_invalid_data=False)
File "linevul_main.py", line 881, in line_level_localization_tp
results = {"total_lines": total_lines,
UnboundLocalError: local variable 'total_lines' referenced before assignment
Hello, what should I do about this error？It appears when running RQ2.Thanks a lot.

[MichaelFu1998-create]

Hi @664730
I'm not sure about the cause given your current error msg, could you show me the complete ``line_level_localization_tp'' function that you used? Thanks!

[664730]

Hi @664730 I'm not sure about the cause given your current error msg, could you show me the complete ``line_level_localization_tp'' function that you used? Thanks!

def line_level_localization_tp(flaw_lines: str, tokenizer, model, mini_batch, original_func: str, args, top_k_loc: list, top_k_constant: list, reasoning_method: str, index: int, write_invalid_data: bool):
# function for captum LIG.
def predict(input_ids):
return model(input_ids=input_ids)[0]

def lig_forward(input_ids):
    logits = model(input_ids=input_ids)[0]
    y_pred = 1 # for positive attribution, y_pred = 0 for negative attribution
    pred_prob = logits[y_pred].unsqueeze(-1)
    return pred_prob

flaw_line_seperator = "/~/"
(input_ids, labels) = mini_batch
ids = input_ids[0].detach().tolist()
all_tokens = tokenizer.convert_ids_to_tokens(ids)
all_tokens = [token.replace("Ġ", "") for token in all_tokens]
all_tokens = [token.replace("ĉ", "Ċ") for token in all_tokens]
original_lines = ''.join(all_tokens).split("Ċ")

# flaw line verification
# get flaw tokens ground truth
flaw_lines = get_all_flaw_lines(flaw_lines=flaw_lines, flaw_line_seperator=flaw_line_seperator)
flaw_tokens_encoded = encode_all_lines(all_lines=flaw_lines, tokenizer=tokenizer)
verified_flaw_lines = []
do_explanation = False
for i in range(len(flaw_tokens_encoded)):
    encoded_flaw = ''.join(flaw_tokens_encoded[i])
    encoded_all = ''.join(all_tokens)
    if encoded_flaw in encoded_all:
        verified_flaw_lines.append(flaw_tokens_encoded[i])
        do_explanation = True

Of course, it's the original one in your code.