Support both pydantic 1.x and 2.x #37055

alanwguo · 2023-07-03T20:33:36Z

Why are these changes needed?

These changes fixes a couple of things when ray and pydantic 2.0 is installed:

ray can start up now
ray jobs works now
gets rid of some deprecation warnings when using old pydantic field names.

Some things are still broken:

ray serve doesn't work because it depends on fastapi and fastapi does not support pydantic 2.0 yet: Prepare for Pydantic V2 release fastapi/fastapi#6051
Pickling pydantic 2 models is not possible, although users can still pickle pydantic 1.0 models via the compatibility pydantic.v1 import.

Related issue number

Checks

I've signed off every commit(by using the -s flag, i.e., git commit -s) in this PR.
I've run scripts/format.sh to lint the changes in this PR.
I've included any doc changes needed for https://docs.ray.io/en/master/.
- I've added any new APIs to the API Reference. For example, if I added a
  method in Tune, I've added it in doc/source/tune/api/ under the
  corresponding .rst file.
I've made sure the tests are passing. Note that there might be a few flaky tests, see the recent failures at https://flakey-tests.ray.io/
Testing Strategy
- Unit tests
- Release tests
- This PR is not tested :(

Signed-off-by: Alan Guo <aguo@anyscale.com>

rkooo567 · 2023-07-06T00:12:54Z

Pickling pydantic 2 models is not possible, although users can still pickle pydantic 1.0 models via the compatibility pydantic.v1 import.

Is this required for ray serve?

Btw, maybe we need a guard against ray serve (or at least warning)? cc @edoakes @sihanwang41

rkooo567

nit comments

python/ray/serve/_private/version.py

rkooo567 · 2023-07-06T00:16:55Z

python/ray/util/state/common.py

 except ImportError:
    # pydantic is not available in the dashboard.
    # We will use the dataclass from the standard library.
    from dataclasses import dataclass

    JobDetails = object
+    is_pydantic_2 = False


why don't we just call it pydantic_1_available? This seems to be more clear with code below

that naming doesn't quite work, because when pydantic is not installed at all, also want init=True. Setting the variable pydantic_1_available = True is weird.

rkooo567 · 2023-07-06T00:19:13Z

python/ray/util/serialization_addons.py

    )

+    if ModelField is not None:


Q: Can't we just check if pydantic is 1 or 2 here instead?

not sure why that's better. I thought it's better to check existance of features/variables over doing a version check.

just the logic seems pretty complicated (1 vs 2 is just one line?) I am okay with this approach too.

in this case, do you think it is possible to add a simple unit test?

shawnpanda · 2023-07-06T16:34:29Z

General question: does this mean that ray serve will be broken until fastapi uses pydantic v2?

rkooo567 · 2023-07-06T22:23:31Z

Btw, this is a release blocker right? Are we tracking the issue @alanwguo ?

alanwguo · 2023-07-06T23:15:20Z

Btw, this is a release blocker right? Are we tracking the issue @alanwguo ?

We pinned pydantic to <2 so this is not a release blocker. But I am still targetting to get this merged for 2.6

alanwguo · 2023-07-06T23:15:44Z

General question: does this mean that ray serve will be broken until fastapi uses pydantic v2?

yes, but only if user is using pydanatic 2.0

Signed-off-by: Alan Guo <aguo@anyscale.com>

rkooo567

I perfer to have an unit test for serialization_addon.py logic (it seems pretty complicated to me). maybe you can do it using runtime _env (runtime_env with pydantic 1 vs 2). Other than that it lgtm

rkooo567 · 2023-07-07T01:27:32Z

python/ray/util/serialization_addons.py

    )

+    if ModelField is not None:


just the logic seems pretty complicated (1 vs 2 is just one line?) I am okay with this approach too.

rkooo567 · 2023-07-07T01:28:35Z

python/ray/util/serialization_addons.py

    )

+    if ModelField is not None:


in this case, do you think it is possible to add a simple unit test?

Signed-off-by: Alan Guo <aguo@anyscale.com>

alanwguo · 2023-07-07T05:37:42Z

I perfer to have an unit test for serialization_addon.py logic (it seems pretty complicated to me). maybe you can do it using runtime _env (runtime_env with pydantic 1 vs 2). Other than that it lgtm

Added a unit test! Verified the test fails if i comment out my serialization_addons.py stuff

rkooo567 · 2023-07-07T08:10:26Z

Lint failure + can you remove the pin from the PR?

alanwguo · 2023-07-07T16:47:53Z

Lint failure + can you remove the pin from the PR?

I don't know if we should remove the pin so quickly. There are still issues with Serve and we can't pickle pydantic models with v 2.x.

@pcmoritz thoughts?

Signed-off-by: Alan Guo <aguo@anyscale.com>

rkooo567 · 2023-07-11T23:35:27Z

Lmk when this is ready to be merged!

alanwguo · 2023-07-12T00:08:48Z

Lmk when this is ready to be merged!

I think it's ready... it's hard to tell with all these test failures, but I looked through them and they look unrelated

Different tests seem to fail between each run.

rkooo567 · 2023-07-12T15:16:47Z

@alanwguo before we merge, is it possible to run CI with pydantic version > 2.0? And then let's merge this by today?

edoakes · 2023-07-12T16:08:34Z

@alanwguo looks like fastapi support pydantic 2.0.3:
https://github.com/tiangolo/fastapi/blob/f7e3559bd5997f831fb9b02bef9c767a50facbc3/pyproject.toml#L45

alanwguo · 2023-07-12T18:56:18Z

@edoakes , thanks for bringing that to my attention!

I don't think I'm the right person to get serve working with pydantic 2.0,
@akshay-anyscale , can we have someone working on serve pick up the remaining work here?

rkooo567 · 2023-07-13T20:43:20Z

The failing tests seem new to me, so I am rerunning it just in case. Highly likely it is not releated

alanwguo · 2023-07-14T18:18:39Z

The failing tests seem new to me, so I am rerunning it just in case. Highly likely it is not releated

Looks ready to merge?

These changes fixes a couple of things when ray and pydantic 2.0 is installed: ray can start up now ray jobs works now gets rid of some deprecation warnings when using old pydantic field names. Some things are still broken: ray serve doesn't work because it depends on fastapi and fastapi does not support pydantic 2.0 yet: Prepare for Pydantic V2 release fastapi/fastapi#6051 Pickling pydantic 2 models is not possible, although users can still pickle pydantic 1.0 models via the compatibility pydantic.v1 import.

These changes fixes a couple of things when ray and pydantic 2.0 is installed: ray can start up now ray jobs works now gets rid of some deprecation warnings when using old pydantic field names. Some things are still broken: ray serve doesn't work because it depends on fastapi and fastapi does not support pydantic 2.0 yet: Prepare for Pydantic V2 release fastapi/fastapi#6051 Pickling pydantic 2 models is not possible, although users can still pickle pydantic 1.0 models via the compatibility pydantic.v1 import. Signed-off-by: NripeshN <nn2012@hw.ac.uk>

These changes fixes a couple of things when ray and pydantic 2.0 is installed: ray can start up now ray jobs works now gets rid of some deprecation warnings when using old pydantic field names. Some things are still broken: ray serve doesn't work because it depends on fastapi and fastapi does not support pydantic 2.0 yet: Prepare for Pydantic V2 release fastapi/fastapi#6051 Pickling pydantic 2 models is not possible, although users can still pickle pydantic 1.0 models via the compatibility pydantic.v1 import. Signed-off-by: e428265 <arvind.chandramouli@lmco.com>

kerthcet · 2023-09-16T03:34:16Z

I don't think I'm the right person to get serve working with pydantic 2.0, @akshay-anyscale , can we have someone working on serve pick up the remaining work here?

Sorry for the ping, any plan to support pydantic 2.0 on ray serve?

kerthcet · 2023-09-16T03:37:32Z

Found the answer, targeting for release 2.8, sorry for the bothering.

These changes fixes a couple of things when ray and pydantic 2.0 is installed: ray can start up now ray jobs works now gets rid of some deprecation warnings when using old pydantic field names. Some things are still broken: ray serve doesn't work because it depends on fastapi and fastapi does not support pydantic 2.0 yet: Prepare for Pydantic V2 release fastapi/fastapi#6051 Pickling pydantic 2 models is not possible, although users can still pickle pydantic 1.0 models via the compatibility pydantic.v1 import. Signed-off-by: Victor <vctr.y.m@example.com>

Support both pydantic 1.x and 2.x

71b30ba

Signed-off-by: Alan Guo <aguo@anyscale.com>

alanwguo force-pushed the support-pydantic-2 branch from d3212fa to 71b30ba Compare July 3, 2023 21:55

alanwguo mentioned this pull request Jul 5, 2023

Cannot Initialize Ray #37019

Closed

alanwguo added 2 commits July 5, 2023 16:45

support serialization of pydantic.v1 models

9dbdad3

Signed-off-by: Alan Guo <aguo@anyscale.com>

fix lint

bcf9a32

Signed-off-by: Alan Guo <aguo@anyscale.com>

alanwguo marked this pull request as ready for review July 5, 2023 23:55

alanwguo assigned rickyyx, rkooo567, simon-mo and pcmoritz Jul 5, 2023

rkooo567 reviewed Jul 6, 2023

View reviewed changes

rkooo567 added the @author-action-required The PR author is responsible for the next step. Remove tag to send back to the reviewer. label Jul 6, 2023

fixup

8454ca2

Signed-off-by: Alan Guo <aguo@anyscale.com>

rkooo567 approved these changes Jul 7, 2023

View reviewed changes

add unit test

e127789

Signed-off-by: Alan Guo <aguo@anyscale.com>

This was referenced Jul 7, 2023

[Core] Issue with installation of ray #36904

Closed

[core] cluster launcher test failure #37060

Closed

alanwguo added 2 commits July 7, 2023 09:48

fix lint

d9f01a8

Signed-off-by: Alan Guo <aguo@anyscale.com>

fix jobs compatibility test

7cd4558

Signed-off-by: Alan Guo <aguo@anyscale.com>

shrekris-anyscale mentioned this pull request Jul 10, 2023

[Ray Clusters] Failed to start Job Supervisor actor: module 'pydantic.fields' has no attribute 'ModelField' #37239

Closed

Merge branch 'master' into support-pydantic-2

a7b6b2f

alanwguo removed the @author-action-required The PR author is responsible for the next step. Remove tag to send back to the reviewer. label Jul 11, 2023

Merge branch 'master' into support-pydantic-2

414d04d

alanwguo force-pushed the support-pydantic-2 branch from 01e5a94 to cacc796 Compare July 12, 2023 18:58

edoakes mentioned this pull request Jul 12, 2023

[serve] Investigate if serve/fastapi works with pydantic>=2.0 #37372

Closed

alanwguo force-pushed the support-pydantic-2 branch from cacc796 to 414d04d Compare July 12, 2023 21:34

rkooo567 merged commit 965ddf3 into ray-project:master Jul 14, 2023

alanwguo mentioned this pull request Jul 17, 2023

Support both pydantic 1.x and 2.x #37468

Merged

8 tasks

akshay-anyscale mentioned this pull request Jul 21, 2023

Add service deployment instructions to stable diffusion template #37645

Closed

8 tasks

massonix mentioned this pull request Aug 22, 2023

export_pseudobulk [BUG] aertslab/pycisTopic#55

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Support both pydantic 1.x and 2.x #37055

Support both pydantic 1.x and 2.x #37055

alanwguo commented Jul 3, 2023 •

edited

Loading

rkooo567 commented Jul 6, 2023

rkooo567 left a comment

rkooo567 Jul 6, 2023

alanwguo Jul 6, 2023

rkooo567 Jul 6, 2023

alanwguo Jul 6, 2023

rkooo567 Jul 7, 2023

rkooo567 Jul 7, 2023

shawnpanda commented Jul 6, 2023

rkooo567 commented Jul 6, 2023

alanwguo commented Jul 6, 2023

alanwguo commented Jul 6, 2023

rkooo567 left a comment

rkooo567 Jul 7, 2023

rkooo567 Jul 7, 2023

alanwguo commented Jul 7, 2023

rkooo567 commented Jul 7, 2023

alanwguo commented Jul 7, 2023

rkooo567 commented Jul 11, 2023

alanwguo commented Jul 12, 2023 •

edited

Loading

rkooo567 commented Jul 12, 2023

edoakes commented Jul 12, 2023

alanwguo commented Jul 12, 2023

rkooo567 commented Jul 13, 2023

alanwguo commented Jul 14, 2023

kerthcet commented Sep 16, 2023

kerthcet commented Sep 16, 2023

Support both pydantic 1.x and 2.x #37055

Support both pydantic 1.x and 2.x #37055

Conversation

alanwguo commented Jul 3, 2023 • edited Loading

Why are these changes needed?

Related issue number

Checks

rkooo567 commented Jul 6, 2023

rkooo567 left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

shawnpanda commented Jul 6, 2023

rkooo567 commented Jul 6, 2023

alanwguo commented Jul 6, 2023

alanwguo commented Jul 6, 2023

rkooo567 left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

alanwguo commented Jul 7, 2023

rkooo567 commented Jul 7, 2023

alanwguo commented Jul 7, 2023

rkooo567 commented Jul 11, 2023

alanwguo commented Jul 12, 2023 • edited Loading

rkooo567 commented Jul 12, 2023

edoakes commented Jul 12, 2023

alanwguo commented Jul 12, 2023

rkooo567 commented Jul 13, 2023

alanwguo commented Jul 14, 2023

kerthcet commented Sep 16, 2023

kerthcet commented Sep 16, 2023

alanwguo commented Jul 3, 2023 •

edited

Loading

alanwguo commented Jul 12, 2023 •

edited

Loading