Generating max_num_tokens.csv for Different Hardware Environments #30

In the part of the experimental model for finding the optimal placement strategy, there are three files in the simdistserve/estimators/profile_data directory, one of which is named max_num_tokens.csv. I am very curious about how this file is generated in different environments, because the other two JSON files can be generated based on the code in evaluation/0-test-single-forward-performance. However, I have not found a method for generating this CSV file. Based on my understanding, this file should be associated with different hardware or GPU environments, and the profiles in different environments should be different. I am eagerly awaiting your response. Thank you very much!