[BugFix] Keep multi-discrete masks boolean when casting specs

[mottopanikeiku]

Description

MultiCategorical.to and MultiOneHot.to cast their masks to the requested sample dtype. The new specs then reject those masks because they are no longer boolean. I now move each mask only to dest_device. The requested sample dtype and the existing device-conversion behavior are unchanged. I also documented that masks stay boolean during dtype conversion.

A small reproducer:

import torch
from torchrl.data import MultiCategorical, MultiOneHot

mask = torch.tensor([False, True, False, True, False])
for cls in (MultiCategorical, MultiOneHot):
    try:
        spec = cls([3, 2], mask=mask).to(torch.int32)
        print(cls.__name__, spec.rand().tolist(), spec.mask.dtype)
    except ValueError as err:
        print(cls.__name__, type(err).__name__, str(err))

	main	This PR
MultiCategorical.to(torch.int32)	ValueError Only boolean masks are accepted.	[1, 0] torch.bool
MultiOneHot.to(torch.int32)	ValueError Only boolean masks are accepted.	[0, 1, 0, 1, 0] torch.bool

Tests:

• Both cases of TestTo.test_multi_masked_dtype fail on main. The test checks exact allowed actions, sample dtype, mask dtype, and that the original spec is unchanged.
• With the fix, test/test_specs.py: 1822 passed, 9 skipped.
• pre-commit passes on the changed files.
• I tested the Python sources without building the optional C++ extension; these spec operations do not use it.

Motivation and Context

Changing the tensor dtype of a valid masked action spec should not invalidate its boolean action mask.

close #4589

• I have raised an issue to propose this change (required for new features and bug fixes)

Types of changes

• Bug fix (non-breaking change which fixes an issue)

Checklist

• I have read the CONTRIBUTION guide (required)
• My change requires a change to the documentation.
• I have updated the tests accordingly (required for a bug fix or a new feature).
• I have updated the documentation accordingly.

I used AI assistance (Claude and GPT-6 Astra) to find, fix and test this bug.

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/rl/4591

• 📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

⚠️ 16 Awaiting Approval

As of commit 11dfb8b with merge base e7a05a7 ():

AWAITING APPROVAL - The following workflows need approval before CI can run:

• Build Aarch64 Linux Wheels (gh)
• Build Linux Wheels (gh)
• Build M1 Wheels (gh)
• Build Windows Wheels (gh)
• Continuous Benchmark (PR) (gh)
• Examples Tests on Linux (gh)
• Generate documentation (gh)
• Habitat Tests on Linux (gh)
• Libs Tests on Linux (gh)
• Lint (gh)
• LLM Tests on Linux (gh)
• Push Binary Nightly (gh)
• SOTA Tests on Linux (gh)
• Tutorials Tests on Linux (gh)
• Unit-tests on Linux (gh)
• Validate Test Partitioning (gh)

This comment was automatically generated by Dr. CI and updates every 15 minutes.