切片索引必须为整数或无，或者具有__index__方法？

Question

#读入数据
with gzip.open('morethan_mfcc.pkl', 'rb') as ifp:
    df = pickle.load(ifp)
pd.set_option('display.width', 100)

df = df.reset_index()
#print(df)
df.drop('index', inplace=True, axis=1)

for index, row in df.iterrows():
     df.at[index, 'lengths']=int(row['MFCC'].shape[0])
print(df.head())
                           MFCC  ........................ lengths
0  [[-10.494777560823318, 19.205118287686812, 16....  ...  1099.0
1  [[-11.632857909892738, 5.8760442872698695, 29....  ...  1099.0
2  [[-8.773702814090234, -12.86215112148708, -7.2...  ...   299.0
3  [[-8.411895008114222, -1.9307392294278005, -7....  ...   299.0
4  [[-6.969186995608002, -17.02179303376442, -23....  ...   299.0

le = LabelEncoder()
le.fit(df['Label'])
le.classes_
print(le.classes_)

#数据预准备
df_train,df_test = train_test_split(df, test_size=0.2, random_state=0)
df_test, df_val = train_test_split(df_test, test_size=0.75, random_state=0)
print('训练数：',df_train.shape[0])
print('验证数：',df_val.shape[0])
print('测试数：',df_test.shape[0])

训练数： 25
验证数： 6
测试数： 1
class BucketedDataIterator():
    def __init__(self, df, num_buckets=7):
        df = df.sort_values('lengths').reset_index(drop=True)
        self.size = len(df) / num_buckets
        self.dfs = []
        for bucket in range(num_buckets):
            self.dfs.append(df.loc[bucket * self.size: (bucket + 1) * self.size - 1])
        self.num_buckets = num_buckets

        # cursor[i] will be the cursor for the ith bucket
        self.cursor = np.array([0] * num_buckets)
        self.shuffle()

        self.epochs = 0

    def shuffle(self):
        # sorts dataframe by sequence length, but keeps it random within the same length
        for i in range(self.num_buckets):
            self.dfs[i] = self.dfs[i].sample(frac=1).reset_index(drop=True)
            self.cursor[i] = 0

    def next_batch(self, n):
        if np.any(self.cursor + n + 1 > self.size):
            self.epochs += 1
            self.shuffle()

        i = np.random.randint(0, self.num_buckets)

        res = self.dfs[i].loc[self.cursor[i]:self.cursor[i] + n - 1]
        self.cursor[i] += n

        # Pad sequences with 0s so they are all the same length
        maxlen = int(max(res['lengths']))

        x = np.zeros([n, maxlen], dtype=np.int32)
        #原码是：x = np.zeros([n, maxlen，32], dtype=np.float)也不行
        for i, x_i in enumerate(x):
            x_i[:res['lengths'].values[i]] = res['MFCC'].values[i]

        return x, le.transform(res['Label'].values), res['lengths']
tr = BucketedDataIterator(df_train)
mini_batch = tr.next_batch(128)
X = mini_batch[0]
y = mini_batch[1]
seqlen = mini_batch[2]
print('X shape:', X.shape)
print('Target:', y[:5])
print('Sequence length:\n', seqlen[:5])


Traceback (most recent call last):
  File "D:/Project/pycharm/RNN/create_mfcc.py", line 164, in <module>
    mini_batch = tr.next_batch(128)
  File "D:/Project/pycharm/RNN/create_mfcc.py", line 160, in next_batch
    x_i[:res['lengths'].values[i]] = res['MFCC'].values[i]
TypeError: slice indices must be integers or None or have an __index__ method

Answer 1

lengths 这个字段你是不是没按我写的转成整数