patrick-kidger / patrick-kidger/jaxtyping

cloudpickle + weakref + index_variadic

Open
#332 4 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
1.9k
Forks
96
PR merge metrics
No merged PRs in 30d

Description

Cloudpickle (latest, 3.3.1 at time of writing) fails to pickle jaxtyped functions because of the weakref (introduced in 0.2.35 of jaxtyping).

I have a MWE using uv with inline packages:

# /// script
# requires-python = ">=3.12"
# dependencies = [
#     "beartype",
#     "cloudpickle==3.1.1",
#     "jaxtyping==0.2.35",
#     "numpy",
# ]
# ///
import beartype
import cloudpickle
import numpy as np
from jaxtyping import Float, jaxtyped


@jaxtyped(typechecker=beartype.beartype)
def typechecked_fn(x: Float[np.ndarray, " d"]) -> float:
    return np.sum(x).item()


def main():
    dumped = cloudpickle.dumps(typechecked_fn)
    print(dumped)
    fn = cloudpickle.loads(dumped)
    print(fn(np.array([1.0, 2.0])))


if __name__ == "__main__":
    main()

When I run this with uv run scratch.py I get

Traceback (most recent call last):
  File "/users/PAS1576/samuelstevens/projects/saev/scratch.py", line 29, in <module>
    main()
  File "/users/PAS1576/samuelstevens/projects/saev/scratch.py", line 22, in main
    dumped = cloudpickle.dumps(typechecked_fn)
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/users/PAS1576/samuelstevens/.cache/uv/environments-v2/scratch-f0adbe6995ced415/lib/python3.12/site-packages/cloudpickle/cloudpickle.py", line 1537, in dumps
    cp.dump(obj)
  File "/users/PAS1576/samuelstevens/.cache/uv/environments-v2/scratch-f0adbe6995ced415/lib/python3.12/site-packages/cloudpickle/cloudpickle.py", line 1303, in dump
    return super().dump(obj)
           ^^^^^^^^^^^^^^^^^
TypeError: cannot pickle 'weakref.ReferenceType' object

When I update jaxtyping to 0.3.2 (in the script metadata)

# /// script
# requires-python = ">=3.12"
# dependencies = [
#     "beartype",
#     "cloudpickle==3.1.1",
#     "jaxtyping==0.3.2",
#     "numpy",
# ]
# ///
import beartype
import cloudpickle
import numpy as np
from jaxtyping import Float, jaxtyped


@jaxtyped(typechecker=beartype.beartype)
def typechecked_fn(x: Float[np.ndarray, " d"]) -> float:
    return np.sum(x).item()


def main():
    dumped = cloudpickle.dumps(typechecked_fn)
    print(dumped)
    fn = cloudpickle.loads(dumped)
    print(fn(np.array([1.0, 2.0])))


if __name__ == "__main__":
    main()

I get the same error:

Traceback (most recent call last):
  File "/users/PAS1576/samuelstevens/projects/saev/scratch.py", line 29, in <module>
    main()
  File "/users/PAS1576/samuelstevens/projects/saev/scratch.py", line 22, in main
    dumped = cloudpickle.dumps(typechecked_fn)
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/users/PAS1576/samuelstevens/.cache/uv/environments-v2/scratch-f0adbe6995ced415/lib/python3.12/site-packages/cloudpickle/cloudpickle.py", line 1537, in dumps
    cp.dump(obj)
  File "/users/PAS1576/samuelstevens/.cache/uv/environments-v2/scratch-f0adbe6995ced415/lib/python3.12/site-packages/cloudpickle/cloudpickle.py", line 1303, in dump
    return super().dump(obj)
           ^^^^^^^^^^^^^^^^^
TypeError: cannot pickle 'weakref.ReferenceType' object

This MWE works on jaxtyping 0.2.34, but my other script fails on 0.2.34 with the same error as #198.

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Reproduce the failure with the inline-package MWE in scratch.py, starting at main() and the cloudpickle.dumps(typechecked_fn) call. Compare the jaxtyping 0.2.34, 0.2.35, and 0.3.2 behaviors around the @jaxtyped function and weakref; done means the function can be dumped and loaded successfully and still returns the expected value.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
devtools
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.