A problem occurs in TNCO env

[spicywei]

When N>1500, the existing code overflows and results in inf Tried to fix it by increasing the temp_power variable so that it can calculate a larger range of floating point numbers, but it didn't work.

[spicywei]

Thank you @Yonv1943 for the help! The original limit of N is 1024, and the maximum maxpower is automatically set with max - 96 by max_tmp_power = int(mult_pow_timess.max().item() - 960), 90 being a value slightly less than the limit of 1024. The problem was then solved by computing the large number with Python's native math.log10(int). Currently can support N up to 20,000.

[Yonv1943]

如果有时间，请帮忙把标题加上方便以后搜索的信息。例如： Computing large numbers dynamically and adaptively。

我没有修改标题的权限 @spicywei

[spicywei]

Ok, thanks for the advice!