以下兩個循環將使用內存，直到我用完，但我不明白為什么。我在每次迭代結束時刪除所有創建的變量，但它仍然泄漏。!pip3 install cupy-cuda101import cupy as cpimport numpy as npfrom sklearn.preprocessing import PolynomialFeaturesxtrain = cp.asnumpy(cp.random.uniform(-1,1,size = (150000,50)))for i in range(0,1000):   weights = cp.random.uniform(-1,1,size = (1275,1000))      for chunk in range(0,xtrain.shape[0],5000):      xchunk = xtrain[chunk:chunk+5000,:]      poly=PolynomialFeatures(interaction_only = True, include_bias = False)      xchunk = cp.array(poly.fit_transform(xchunk))      ranks = cp.matmul(xchunk,weights)      del ranks, xchunk, poly   del weightsxtrain 也只是浮點數據，介于 -1 和 1 之間。