[Serialization] Update CloudPickle to 1.6.0#9694
[Serialization] Update CloudPickle to 1.6.0#9694rkooo567 merged 2 commits intoray-project:masterfrom
Conversation
|
Can one of the admins verify this patch? |
a5d4483 to
05a759c
Compare
84b77e1 to
4a19fb7
Compare
|
Test FAILed. |
|
Test FAILed. |
|
Test FAILed. |
|
retriggered build to use the latest upstream |
4a19fb7 to
6ae637e
Compare
|
retriggered build to use the latest upstream |
|
Test FAILed. |
|
It seems |
|
@suquark Do you have any plan to take a look at this sooner or later? |
6ae637e to
66f2c3e
Compare
|
let's see if rebasing would result in a successful build |
|
Still fails on |
|
not sure what actually happened. the test is green on my local macbook. I may not have enough time to fully investigate into this issue recently. |
|
No problem! Please take ur time. If you can point me resources, I can take a look if I can fix the issue |
|
It only seems like the macos run takes more time than the linux one. I do not find any problem on my own macbook |
|
Ah, I see. Our Mac test is usually slower than Linux, and it causes timeout error for "small" bazel tests (which is supposed to be done within 60 seconds I believe). Just move the test to medium size test Line 84 in 0d22c0b |
efba55b to
18f8e4c
Compare
This reverts commit f0c3910.
* update cloudpickle to 1.6.0 * fix CI timeout
Why are these changes needed?
After #9622, we no longer need the numpy workaround in cloudpickle. Also the latest cloudpickle natively supports pickle5. This PR ships the latest cloudpickle to Ray without any patch (except adding "ray" prefix in
__init__.pyfor correct module import)We also move some serialization tests to the correct test script.
Related issue number
Checks
scripts/format.shto lint the changes in this PR.