First of all, I want to start by saying thank you for this great work. However I found a few issues that do not let me successfully execute the LogPPT technique

1. After executing "fewshot_sampling.py", the label files are generated using the corrected Oracle files, provided by the paper "Guidelines for Assessing the Accuracy of Log Message Template Identification Techniques", and they are required to successfully run the train() function. My question is, how one can use the dataset, whose corrected Oracle files are not available? or even, the Oracle files are not available at all?. Of course one option is to use any other log parser to generate templates and use it as oracle, but it is not a feasible solution.

2. Although the benchmark.py does not do parsing at all, and basically calculate the log parsing accuracy metrics value based on the provided ground truth and parsed files in your repository (in the output folder). Still, the results provided in the paper and local execution does not match. Is there any other version of datasets that were utilized?. The similar issue is mentioned in another issue by another user ( link of the issue is here )

Could you kindly help me with the issues (specifically the first issue)