Skip to content

chore(deps): bump the inference-dependencies group across 1 directory with 6 updates #525

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
dependabot wants to merge 1 commit into from
Closed

chore(deps): bump the inference-dependencies group across 1 directory with 6 updates #525

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
12 changes: 6 additions & 6 deletions evaluation/pyproject.toml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4,25 +4,25 @@ version = "0.1.0"
description = "Evaluation workflow runtime dependencies"
requires-python = ">=3.12"
dependencies = [
"numpy==2.2.6",
"numpy==2.4.4",

@github-actions github-actions Bot Apr 21, 2026

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

numpy 2.2.6 → 2.4.4 — ABI-sensitive, stays within 2.x

Surface: python-runtime (ABI-sensitive per rubric). Source: numpy 2.4.4 release notes.

This bump stays within the NumPy 2.x series (2.2 → 2.4); no ABI break is expected between these minor versions. The 2.4.x releases are patch-level fixes: OpenBLAS threading on ARM (issue #30816), FNV-1a 64-bit hash fix, ufunc where=True warning.

Note: training/rl/pyproject.toml and training/il/lerobot/pyproject.toml both pin numpy==1.26.4 (NumPy 1.x). Those surfaces are not touched by this PR, so there is no cross-surface numpy ABI conflict introduced here. Validate with ruff check and targeted pytest in evaluation/.

"azure-core==1.39.0",
"azure-storage-blob==12.28.0",
"azure-identity==1.25.3",
"azure-ai-ml==1.32.0",
"marshmallow==3.26.2",
"marshmallow==4.3.0",

@github-actions github-actions Bot Apr 21, 2026

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

⚠️ marshmallow 3.26.2 → 4.3.0 — major-version breaking changes

Source: marshmallow CHANGELOG (4.0.0 entry — see "upgrading_4_0").

Key backwards-incompatible changes since 3.x:

  • @validates-decorated methods now receive data_key as a keyword argument (was positional).
  • fields.UUID no longer subclasses fields.String.
  • Schema.load() no longer silently skips schema validators when a generator is passed.
  • from_iso_date, from_iso_time, from_iso_datetime, isoformat, to_iso_time, to_iso_datetime, from_rfc, rfcformat, is_keyed_tuple, get_fixed_timezone removed from marshmallow.utils.
  • fields.Boolean no longer serializes non-boolean values.
  • Date/time deserialization now uses the standard-library fromisoformat methods.
  • Security fix included: CVE-2025-68480 was fixed in marshmallow 4.1.2 (merged error store message handling). Source: CHANGELOG.rst 4.1.2 entry.

If any evaluation code calls the removed utility functions or relies on the old @validates signature, it will raise at runtime.

"mlflow==3.11.1",
"packaging==26.1",
"psutil==7.2.2",
"pynvml==13.0.1",
"pyperclip==1.11.0",
"onnx==1.21.0",
"onnxscript==0.6.2",
"onnxscript==0.7.0",

@github-actions github-actions Bot Apr 21, 2026

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

onnxscript 0.6.2 → 0.7.0 — minor bump, new GraphBuilder API

Source: onnxscript v0.7.0 release notes.

Notable additions: new onnxscript.nn module (Module, Parameter, Sequential), new GraphBuilder API for programmatic ONNX graph construction, rewriter/optimizer improvements (commutative-ops expansion, BatchNorm+Conv fusion fixes, DynamicQuantizeLinear constant-folding guard). No documented breaking changes. Low risk for this surface.

"onnxruntime-gpu==1.24.4",
"toml==0.10.2",
"gymnasium==1.2.3",
"torch==2.10.0",
"tensordict==0.12.1",
"lerobot==0.5.0",
"torch==2.11.0",

@github-actions github-actions Bot Apr 21, 2026

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

⚠️ torch 2.10.0 → 2.11.0 — CUDA breaking changes

Two backwards-incompatible GPU changes in this release (full release notes):

  1. PyPI wheels now bundle CUDA 13.0 (pip install torch on PyPI now installs CUDA 13.0 wheels instead of CUDA 12.x). Hosts with only CUDA 12.x drivers will fail to load the installed wheel. Use --index-url (download.pytorch.org/redacted) to pin CUDA 12.8, or cu126` for CUDA 12.6.

  2. Volta (SM 7.0 / V100) support removed from CUDA 12.8 and 12.9 builds. If any evaluation node runs on V100 GPUs and uses CUDA 12.8+, those nodes will fail. Use the CUDA 12.6 build URL (cu126) to retain V100 support.

Recommended: Verify the CUDA driver version and GPU architecture of evaluation nodes before merging. Run ruff check and a GPU smoke test via the owning package's test suite.

"tensordict==0.12.2",

@github-actions github-actions Bot Apr 21, 2026

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

tensordict 0.12.1 → 0.12.2 — ABI-sensitive patch release

Surface: python-runtime (ABI-sensitive per rubric). Source: TensorDict v0.12.2 release notes.

Patch release: fixes _ragged_idx loss during consolidation of nested tensors, which caused numerical incorrectness when the nested tensor had more than 2 dimensions and ragged_idx != 1. No breaking API changes. Low risk on its own, but should be validated alongside the torch==2.11.0 bump since both affect GPU tensor operations.

"lerobot==0.5.1",
]

[build-system]
Expand Down
Loading