Dear Author,
Greetings!
I am currently attempting to reproduce the experimental results on the IMDB dataset from your paper "Lero: A Learning-to-Rank Query Optimizer" and have encountered some difficulties for which I seek your guidance.
First and foremost, I would like to express my admiration for the innovative approach presented in your paper, where you propose replacing the regression problem of predicting latency with a learning-to-rank classification problem. I believe this has significant implications for the field of databases. In my reproduction of your work, I have strictly followed the code and database construction guidelines you provided (referencing the construction of the IMDB database in the paper "Is Your Learned Query Optimizer Behaving As You Expect? A Machine Learning Perspective") and have utilized the pre-trained imdb_pw model.
However, during the execution, I have found that the current runtime significantly exceeds the expected execution time mentioned in Table 1 of your paper. Specifically, my experiment has been running for approximately 26 hours and has not yet completed. This is quite perplexing for me, as such a long runtime poses a considerable challenge to my research schedule.
I would like to inquire whether it is advisable to retrain the model, or if there might be an issue with the version differences in the construction of the IMDB database?
I understand that you may be very busy, but any assistance you can provide would be invaluable to me. If necessary, I can supply more experimental details or error logs to help you better understand the issues I am encountering.
Thank you very much for your time and help. I look forward to your response.
Best regards
Dear Author,
Greetings!
I am currently attempting to reproduce the experimental results on the IMDB dataset from your paper "Lero: A Learning-to-Rank Query Optimizer" and have encountered some difficulties for which I seek your guidance.
First and foremost, I would like to express my admiration for the innovative approach presented in your paper, where you propose replacing the regression problem of predicting latency with a learning-to-rank classification problem. I believe this has significant implications for the field of databases. In my reproduction of your work, I have strictly followed the code and database construction guidelines you provided (referencing the construction of the IMDB database in the paper "Is Your Learned Query Optimizer Behaving As You Expect? A Machine Learning Perspective") and have utilized the pre-trained imdb_pw model.
However, during the execution, I have found that the current runtime significantly exceeds the expected execution time mentioned in Table 1 of your paper. Specifically, my experiment has been running for approximately 26 hours and has not yet completed. This is quite perplexing for me, as such a long runtime poses a considerable challenge to my research schedule.
I would like to inquire whether it is advisable to retrain the model, or if there might be an issue with the version differences in the construction of the IMDB database?
I understand that you may be very busy, but any assistance you can provide would be invaluable to me. If necessary, I can supply more experimental details or error logs to help you better understand the issues I am encountering.
Thank you very much for your time and help. I look forward to your response.
Best regards