代码是关于在gpu上填充每个文本的。
培训了整个项目几天后，出现了一个问题
import sklearn
print('The scikit-learn version is {}.'.format(sklearn.__version__))
X, y = make_classification(n_classes=2, class_sep=0, weights=[0.05,0.95],n_clusters_per_class=2,
n_features=2, n_samples=10000, n_informative=2, n_redundant=0, n_repeated=0)
#Repeated ENN
renn = RepeatedEditedNearestNeighbours( n_neighbors = 5, n_jobs= 2, max_iter = 100)
#Oversampling after have undersampled
smote_enn = SMOTEENN()
#Classifier
classifier = LogisticRegression(random_state = 0)
# Make the splits
n = 2
kf = StratifiedKFold(n_splits = n, random_state = 0)
# Create regularization penalty space
penalty = ['l1', 'l2']
# Create regularization hyperparameter space
C = np.logspace(0, 4, 10)
# Create hyperparameter options
parameters = dict(C=C, penalty=penalty)
random_search = RandomizedSearchCV(pipeline, param_distributions=parameters, n_iter=1000, cv = kf, return_train_score = True)
gg = random_search.fit(X, y)
gg .best_estimator_
random_search.cv_results_
，错误是
x = torch.from_numpy(x.astype(int))
我搜索了该错误的几个答案，但是这些都不是由RuntimeError: Trying to resize storage that is not resizable at /pytorch/aten/src/TH/generic/THStorage.c:183触发的
x = torch.from_numpy(x.astype(int))它运行了好几天，但是突然出错了。是因为我的GPU（Tesla K80）的内存不足？错误 max_seq_len = max(lengths)
batch_size = len(batch)
x = np.zeros((batch_size, max_seq_len))
for i, text_ids in enumerate(texts_ids):
padded = np.zeros(max_seq_len) # padded = [0,0,0,....,max_sqe_len]
padded[:len(text_ids)] = text_ids # padded = [50,4,16,...,0,0,0]
x[i, :] = padded # x = [[text],[text][text]] indexed by batches
x = torch.from_numpy(x.astype(int))
x = move_to_cuda(x)
lengths = move_to_cuda(lengths)
labels = move_to_cuda(labels)
是什么意思？