numpy 和 pytorch 中，对空的 tensor 取 sum 是 0，取 mean 是 NaN！对此，numpy 会报 Runtime Warning，而 pytorch 却没有一丝提示… 对张量用 mask 之后可能会产生空的 tensor（即 mask 全是 False 没有 True），可能之前训练 loss 都正常，突然就 NaN 了。所以取 mean 前先判空。 Input array or object that can be converted to an array. NumPy: Remove rows/columns with missing value (NaN) in ndarray Array containing numbers whose sum is desired. Where the array I'm working on consist of None, which means to ignore that value in the processing. With the argument axis=1, any () tests whether there is at least one True for each row. Syntax: numpy.nanmean(a, axis=None, dtype=None, out=None, keepdims=)) Parameters: a: [arr_like] input array axis: we can use axis=1 means row wise or axis=0 means column wise. This article describes the following contents. However, np.average doesn't ignore NaN like np.nanmean does, so my first 5 entries of each row are included in the latitude averaging and make the entire time series full of NaN. Is there a way I can take a weighted average without the NaN's being included in the calculation? numpy.nansum — NumPy v1.15 Manual numpy.nanmean() function can be used to calculate the mean of array ignoring the NaN value. float64 intermediate and return values are used for integer inputs. New in version 1.9.0. Parameters a array_like. If X is a matrix, then nanmean(X) is a row vector of column means, computed after removing NaN values.. The standard deviation is computed for the flattened array by … np.isnan (arr) Output : [False True False False False False True] The output array has true for the indices which are NaNs in the original array and false for the rest. numpy.nanvar — NumPy v1.24.dev0 Manual nanstd (a, axis=None, dtype=None, out=None, ddof=0, keepdims=