Bump tensordict to 0.12.4 minimum

[peterdsharpe]

PhysicsNeMo Pull Request

Description

TensorDict 0.12.4 includes some new critical torch.compile and thread-safety improvements for these issues, which are relevant to GLOBE, PNM-Mesh, and our unified external aero recipe:

• [BUG] @tensorclass silently skips __post_init__ under torch.compile, producing wrong output pytorch/tensordict#1708
• [BUG] @tensorclass field defaults are not applied under torch.compile, raising KeyError pytorch/tensordict#1710
• [BUG] Tensordict is not Threadsafe for device transfer in all cases pytorch/tensordict#1700

Fixes for these were added here: pytorch/tensordict#1709 and pytorch/tensordict#1711.

And these are now released as part of TensorDict 0.12.4.

Checklist

• I am familiar with the Contributing Guidelines.
• New or existing tests cover these changes.
• The documentation is up to date with these changes.
• The CHANGELOG.md is up to date with these changes.
• An issue is linked to this pull request.
• If I am implementing a new model or modifying any existing model, I have followed the Models Implementation Coding Standards.

Dependencies

Review Process

All PRs are reviewed by the PhysicsNeMo team before merging.

Depending on which files are changed, GitHub may automatically assign a maintainer for review.

We are also testing AI-based code review tools (e.g., Greptile), which may add automated comments with a confidence score.
This score reflects the AI’s assessment of merge readiness and is not a qualitative judgment of your work, nor is
it an indication that the PR will be accepted / rejected.

AI-generated feedback should be reviewed critically for usefulness.
You are not required to respond to every AI comment, but they are intended to help both authors and reviewers.
Please react to Greptile comments with 👍 or 👎 to provide feedback on their accuracy.

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[peterdsharpe]

/blossom-ci

[greptile-apps]

Greptile Summary

This PR removes the <0.12 upper bound on tensordict and sets a new minimum of >=0.12.4, which includes upstream fixes for torch.compile regressions and thread-safety issues relevant to GLOBE, PNM-Mesh, and the unified external aero recipe.

• pyproject.toml: tensordict constraint changed from >=0.11.0,<0.12 to >=0.12.4; the now-resolved TODO comments are also cleaned up.
• uv.lock: Lock file regenerated; beyond tensordict itself, uv also advanced several unrelated packages to their latest compatible versions (e.g. aiobotocore, botocore, click, certifi, numpy 2.4.4→2.4.6), which is expected behaviour when re-solving the full dependency graph.

Important Files Changed

Filename	Overview
pyproject.toml	Bumps tensordict minimum from >=0.11.0,<0.12 to >=0.12.4, removing the torch.compile regression upper bound and its associated comments.
uv.lock	Lock file regenerated after the tensordict bump; many unrelated packages (aiobotocore, botocore, click, certifi, cachetools, numpy, etc.) were also updated to their latest compatible versions as a side effect of running uv lock.

_{Reviews (1): Last reviewed commit: "Bump tensordict to 0.12.4 minimum" | Re-trigger Greptile}

[coreyjadams]

Considering half these PRs are coming to tensordict from our dev team, LGTM 😅

[peterdsharpe]

/blossom-ci

[peterdsharpe]

/ok to test 01bc273

[peterdsharpe]

/ok to test 604c5ec

[peterdsharpe]

/blossom-ci

[peterdsharpe]

/blossom-ci

[peterdsharpe]

/blossom-ci