mindspore将自定义数据集转换为MindRecord格式

问题描述：

电脑是win10 64bit，cpu运行，MindSpore版本1.5.0，python版本3.9。运行的是文档中数据集加载中，数据集转换的代码：

from io import BytesIO
import os
import mindspore.dataset as ds
from mindspore.mindrecord import FileWriter
import mindspore.dataset.vision.c_transforms as vision
from PIL import Image


MINDRECORD_FILE = "test.mindrecord"

if os.path.exists(MINDRECORD_FILE):
    os.remove(MINDRECORD_FILE)
    os.remove(MINDRECORD_FILE + ".db")

writer = FileWriter(file_name=MINDRECORD_FILE, shard_num=1)

cv_schema = {"file_name": {"type": "string"}, "label": {"type": "int32"}, "data": {"type": "bytes"}}
writer.add_schema(cv_schema, "it is a cv dataset")

writer.add_index(["file_name", "label"])

data = []
for i in range(100):
    i += 1

    sample = {}
    white_io = BytesIO()
    Image.new('RGB', (i*10, i*10), (255, 255, 255)).save(white_io, 'JPEG')
    image_bytes = white_io.getvalue()
    sample['file_name'] = str(i) + ".jpg"
    sample['label'] = i
    sample['data'] = white_io.getvalue()

    data.append(sample)
    if i % 10 == 0:
        writer.write_raw_data(data)
        data = []

if data:
    writer.write_raw_data(data)

writer.commit()

最后一步writer.commit()出错，错误代码：

Traceback (most recent call last):
  File "D:\。。。\GenerateTest.py", line 47, in <module>
    writer.commit()
  File "C:\Users\zhuyue\AppData\Local\Programs\Python\Python39\lib\site-packages\mindspore\mindrecord\filewriter.py", line 391, in commit
    self._generator.build()
  File "C:\Users\zhuyue\AppData\Local\Programs\Python\Python39\lib\site-packages\mindspore\mindrecord\shardindexgenerator.py", line 54, in build
    ret = self._generator.build()
RuntimeError: Unexpected error. Failed to open file, file path: D:\。。。\test.mindrecord
Line of code : 66
File         : D:\mindspore\mindspore\ccsrc\minddata\mindrecord\meta\shard_header.cc
Process finished with exit code 1

解答：

1. windows平台上，目前还不支持 MINDRECORD_FILE = "test.mindrecord" 放在中文路径下。

即： 这个路径中不能有中文。