Results 32 comments of 伍侃

@gab23r I have use `cache=True` I think `single thread` is key point

`cumsum_py3` and `cumsum_py4` are hold the GIL, why `cumsum_py4` so slow? they are same code, except the numba function position.

On consumer-grade GPU, VRAM is a big problem, and it will be embarrassing if data is large. It is more appropriate to use `DSL` in the professional field. Currently in...

I find some code can to do that. https://github.com/jamesmontemagno/TextToSpeechPlugin/blob/master/src/TextToSpeech.Plugin/TextToSpeech.apple.cs#L113

``` t1 = BANCHMARKINDEXCLOSE < BANCHMARKINDEXOPEN COUNT((CLOSE > OPEN) & t1, 50) / COUNT(t1, 50) ``` `operator 'truediv' not implemented for bool dtypes` 除法的参数不能为bool类型 COUNT(bool类型的数组, 50),会将bool类型当成1或0进行求和。按理来说COUNT结果应当是np.ndarray类型的整数,而不是纯bool

会不会是一维和二维混用导致? 我这边没办法复现

你用的版本可能过老。 参考readme中,开发人员安装 这一节

https://github.com/wukan1986/ta_cn/blob/main/ta_cn/__init__.py#L5 numba报错时,一般把cache关了就可以了,但你这地方好像并不是这个问题导致。 也有可能是python版本与numba版本问题。 我现在环境中版本为: Package Version ------------------------ --------- Bottleneck 1.3.4 numba 0.55.2 numexpr 2.8.1 numpy 1.22.4 pandas 1.4.1 Python 3.8.3

由于已经有了新的想法,所以,这个项目基本已经不维护了,请使用polars_ta这个项目

下次请pull到dev分支吧