[mindspore] [read graph data] cannot read mindrecord format graph data

Mindspore.DataSet.GraphData

【 Operation steps & Problem phenomenon 】

1、 The original question ： The first time I used this function was in 1.0.0 edition （ stay 1.0.0 The environment installed when it was first released ）, Data can be read normally .

Later, it was changed to 1.0.1 and 1.1.0 Cannot read after version , And return 1.0.0 edition （ reinstall ） It cannot be read later , The problem is 1 month 26 Number appears .

Display error yes

File "C:\Users\15561\.conda\envs\MS\lib\site-packages\mindspore\dataset\engine\graphdata.py", line 81, in __init__

    self._graph_data = GraphDataClient(dataset_file, num_parallel_workers, working_mode, hostname, port)

RuntimeError: Thread ID 1 Unexpected error. Fail to open./data/cora

Line of code : 92

File         : D:\jenkins\agent-working-dir\workspace\Compile_CPU_Windows\mindspore\mindspore\ccsrc\minddata\dataset\engine\gnn\graph_loader.cc

Solutions that have been tried ：

It has been reinstalled conda once , It's reassembled win10 once .

Use open Function reads the path file successfully , and GraphData Cannot read .

The dataset has been regenerated , It is exactly the same as the original data set .

2、 Possible sources of error ：

Github Modification log on ： !11139 [MD][BugFix] MindData remove the compile link of _c_dataengine… · mindspore-ai/[email protected] · GitHub

1. This problem is confirmed , It is our Windows One on the platform bug, The main performance is ：

If in A Generate under directory mindrecord file , And then put mindrecord File move to B Catalog , Based on the B Directory read mindrecord file , Then there will be mistakes .

2. I suggest you avoid ：

Generate mindrecord after , Don't move , Then use it directly .

3. another , We have been repairing this internally bug 了 , After repair , I will explain in this post , Then you can verify again .