Use tempfile for storing temporary checkpoints/params/models

[zlsh80826]

需求描述 Feature Description

目前 Paddle 實現的多個單測中, 會有儲存 checkpoints 或 inference model 後再讀取的測試, 這類型的測試會將模型儲存, 並在完成單測後刪除。但是多個單測會有暫存檔案同名的問題, 會導致單測同時運行時互相覆蓋掉其他單測的暫存檔案, 導致單測會出現隨機性的失敗。例如: test_dataset 和 test_dataset_consistency_inspection 會使用相同 test_run_with_dump_a.txt 作為暫存檔案，又比如 test_resnet 和 test_build_strategy 都使用相同的 resnet.dygraph.params 作為模型暫存檔。

在撰寫單測時應該注重單測的孤立性, 單測前後系統的狀態需要盡量保持相同, 最大限度地減少生成檔案或是測試之間的依賴關係, 如果有需要使用暫存檔案, 可以使用 mktemp 或是 python 模塊的 tempfile, 並在單測的 tearDown 函式中釋放從系統中取得的資源。

替代实现 Alternatives

有用到暫存檔案的單測應該參考類似 test_pass_builder.py 的寫法, 使用 tempfile 模塊底下的函式, 照單測的需求使用對應建立暫存檔案或目錄的 API 來減少單測間的影響。

[paddle-bot-old]

您好，我们已经收到了您的问题，会安排技术人员尽快解答您的问题，请耐心等待。请您再次检查是否提供了清晰的问题描述、复现代码、环境&版本、报错信息等。同时，您也可以通过查看官网API文档、常见问题、历史Issue、AI社区来寻求解答。祝您生活愉快～

Hi! We've received your issue and please be patient to get responded. We will arrange technicians to answer your questions as soon as possible. Please make sure that you have posted enough message to demo your request. You may also check out the API，FAQ，Github Issue and AI community to get the answer.Have a nice day!

[zlsh80826]

#42626 fixes part of the unit tests, but there are still many unit tests should be refactored.

[jzhang533]

cc: @kolinwei

[Aurelius84]

dygraph_to_static目录下的所有单测均已替换为tmpfile来管理，已cherry-pick到 2.3 分支，见：#42860

[zlsh80826]

Thank @Aurelius84, we will check it ASAP.

[luotao1]

we will check it ASAP.

@zlsh80826 How about the check result, could we close this issue?

[zlsh80826]

Hi @luotao1,
https://github.com/PaddlePaddle/Paddle/blob/develop/python/paddle/fluid/tests/unittests/ir/test_convert_to_mixed_precision.py this ut doesn't use tempfile for temporary models/params.

[luotao1]

@jiweibo Could you fix the test_convert_to_mixed_precision.py?

[jiweibo]

@jiweibo Could you fix the test_convert_to_mixed_precision.py?

#48819 will fix the problem.

[zlsh80826]

Hi @luotao1,
dist_pass_test_base.py doesn't use tempfile for temporary files, too. Besides, there is a flaw in dist_pass_test_base.py#L193, the mode argument of os.makedirs should be represent as oct format (reference), which should be 0o777 instead of 777. The incorrect mode=777 will be interpreted as 0o1411 and it will cause permission denied for non-root user. Thanks

[sneaxiy]

Hi @luotao1, dist_pass_test_base.py doesn't use tempfile for temporary files, too. Besides, there is a flaw in dist_pass_test_base.py#L193, the mode argument of os.makedirs should be represent as oct format (reference), which should be 0o777 instead of 777. The incorrect mode=777 will be interpreted as 0o1411 and it will cause permission denied for non-root user. Thanks

@zlsh80826 Thanks for pointing out. We would fix this issue.

[zlsh80826]

Thank all of you,
Once test_dist_pass_test_base and test_convert_to_mixed_precision are solved, I will close the issue.

[sneaxiy]

Hi @luotao1, dist_pass_test_base.py doesn't use tempfile for temporary files, too. Besides, there is a flaw in dist_pass_test_base.py#L193, the mode argument of os.makedirs should be represent as oct format (reference), which should be 0o777 instead of 777. The incorrect mode=777 will be interpreted as 0o1411 and it will cause permission denied for non-root user. Thanks

@zlsh80826 Thanks for pointing out. We would fix this issue.

@zlsh80826 This is fixed by #48863 .

[zlsh80826]

@sneaxiy appreciate your help!

[luotao1]

Once test_dist_pass_test_base and test_convert_to_mixed_precision are solved, I will close the issue.

• test_convert_to_mixed_precision.py use tempfile for temporary models/params #48819 and - Fix tmp directories creation in dist_pass_test_base.py #48863 are both merged, could you close this issue? @zlsh80826

[zlsh80826]

Sure, thanks!