第一列看起来不像是float64。
您可以省略dtype=...，并检查数据的类型：

import pandas as pd 
import numpy as np

df = pd.read_csv('aars.tsv', sep='\t', usecols=['label','sequence'])

for item in df.values:
    for i in range(item.size):
       print(type(item[i]), end=" ")
    print()

这将输出如下所示的内容（当我创建正确的输入时，我添加了一行列标题）：

<class 'str'> <class 'int'>

下面是一个用PandasStringMethods和pandas.Series.explode做的命题：

import pandas as pd

out= (
        pd.read_csv("AARS.tsv", sep="\t", usecols=["sequences"])
            .assign(temp= lambda x: x["sequences"].str.strip("[]")
                                                  .str.replace("\]\s*,\s*\[", ", ",
                                                               regex=True)
                                                  .str.split(","))
                                                  .explode("temp")
                                                  .astype(float)
                                                  .values
     )

#输出：

print(out)

[[0.]
 [1.]
 [0.]
 [0.]
 [0.]
 [0.]
 [1.]
 [0.]
 [0.]
 [0.]
 [1.]
 [0.]
 [0.]
 [0.]
 [1.]
 [0.]
 [0.]
 [0.]
 [1.]
 [0.]
 [0.]
 [0.]
 [1.]
 [0.]]

print(type(out)

numpy.ndarray

如果您需要将数组重新调整为2D，请使用numpy.reshape：

print(np.reshape(out, (-1, 2)))

[[0. 1.]
 [0. 0.]
 [0. 0.]
 [1. 0.]
 [0. 0.]
 [1. 0.]
 [0. 0.]
 [1. 0.]
 [0. 0.]
 [1. 0.]
 [0. 0.]
 [1. 0.]]