This is a little obfuscated:

1. The function's name does not match the expected return value type/format. A check should return True/False.
2. Returning 9e99 is not clear on what is going on.
3. Creating a folder does not confirm the run status. For this test to be true to its purpose, we need to guarantee that it has not finished, not just started.

We should consider using a get_experiment_status function to check if all iterations have finished running; this function would return at least "RUNNING" and "FINISHED", other states might be helpful but are not required now.

1. The function's name does not match the expected return value type/format. A check should return True/False.

Can possibly change it to get_trial_times

2. Returning 9e99 is not clear on what is going on.

It is just a dummy value, to compare for latest modified time in while loop line 121-128

            # Check if first config is complete
            while True:
                cur_modified_time = self.check_trial_times()
                print(f"Current modified time: {cur_modified_time}")
                print(f"Latest modified time: {latest_modified_time}")
                if abs(cur_modified_time - latest_modified_time) < 1e-3:
                    break
                latest_modified_time = cur_modified_time
                time.sleep(10)

3. Creating a folder does not confirm the run status. For this test to be true to its purpose, we need to guarantee that it has not finished, not just started.

This function returns the latest modified time of a given iteration (folder names are matched using iteration glob) - so if a run is completed the folder should no longer be modified.

4, We should consider using a get_experiment_status function to check if all iterations have finished running; this function would return at least "RUNNING" and "FINISHED", other states might be helpful but are not required now.

get_experiment_status is helpful in general, but might not be too useful in resume_check because we need to check iteration completion, as opposed to experiment completion (or all_iterations completion)