如何有效地将 function 应用于大型 pandas 系列？

Question

I have pandas series of time samples with a start date in unix time. Each time sample is x * 1 / 512 so time stamp 0 = 0, time stamp 2 = 1 / 512 or 0.00195, time stamp 3 = 2 / 512 or 0.0039. I need to add the start date (offset) to all values and convert the result to local time (PST). I have the following

times = np.arange(0, 3600, 1/512)
tz = 'US/Pacific'
offset = 1569603352 # 2019-09-27 09:57 (or something similar)
srs = pd.Series(times)
srs.apply(lambda t: pd.to_datetime(offset + t, unit='s', utc=True) \
                      .tz_convert(tz))

is there some way to speed this up? I have a bunch of powerful GPUs and around 50 threads so multithreading or processing is available.

Answer 1

I will speed up by remove apply

pd.to_datetime(offset + srs, unit='s', utc=True).dt.tz_convert(tz)