Skip to content

chore(deps-dev): Bump the python-deps group across 1 directory with 2 updates #33

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-dev): Bump the python-deps group across 1 directory with 2 updates #33

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
14 changes: 7 additions & 7 deletions pyproject.toml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -93,7 +93,7 @@ robosuite = { git = "https://github.com/ARISE-Initiative/robosuite.git", rev = "

[dependency-groups]
dev = [
"pytest>=8,<9", # reason: test runner. Pin <9: pytest 9.0.3 has a collection regression where multiple modules with module-level pytest.importorskip / pytest.skip(allow_module_level=True) emit "ERROR: found no collectors" and conflate every SKIPPED entry under the first module — silently hides skip reasons in tests/sim/ (breaks tests/unit/test_sim_collection.py).
"pytest>=8,<10", # reason: test runner. Pin <9: pytest 9.0.3 has a collection regression where multiple modules with module-level pytest.importorskip / pytest.skip(allow_module_level=True) emit "ERROR: found no collectors" and conflate every SKIPPED entry under the first module — silently hides skip reasons in tests/sim/ (breaks tests/unit/test_sim_collection.py).
"pytest-asyncio>=0.23", # reason: async test support
"pytest-cov>=5", # reason: coverage reports
"hypothesis>=6.100", # reason: schema fuzz testing
Expand Down Expand Up @@ -137,12 +137,12 @@ sim = [
"num2words>=0.5.14",
"pymunk<8",
"robot-descriptions>=1.12.0", # reason: SO-100 MJCF from DeepMind mujoco_menagerie for offscreen render
"transformers>=5.4.0,<5.14.0",
"transformers>=5.4.0,<5.15.0",
"lerobot[diffusion]", # reason: lerobot 0.6.0 moved `diffusers` behind the `diffusion` extra; the Diffusion Policy adapter (diffusion-pusht rSkill) imports lerobot.policies.diffusion which needs it. Bare lerobot (transitive) no longer carries diffusers.
]
libero = [
"lerobot[libero]", # reason: pulls hf-libero (linux-only) + LIBERO bddl task files + robosuite
"transformers>=5.4.0,<5.14.0", # reason: lerobot 0.6.0 requires transformers>=5.4.0,<5.14.0 via its `transformers-dep` extra (smolvla / pi / xvla / libero all alias it). Stated explicitly here — matching the sim / metaworld / maniskill3 / robocasa groups — so a multi-group sync resolves to one version and the pin survives even if a future lerobot loosens its own bound.
"transformers>=5.4.0,<5.15.0", # reason: lerobot 0.6.0 requires transformers>=5.4.0,<5.15.0 via its `transformers-dep` extra (smolvla / pi / xvla / libero all alias it). Stated explicitly here — matching the sim / metaworld / maniskill3 / robocasa groups — so a multi-group sync resolves to one version and the pin survives even if a future lerobot loosens its own bound.
"num2words>=0.5.14", # reason: required by SmolVLAProcessor (transformers) for number tokenization
"bitsandbytes>=0.45", # reason: 4-bit (NF4) quantization for π0.5 — 3.4 B PaliGemma backbone needs ≤8 GiB VRAM in nf4
"accelerate>=1.13.0", # reason: init_empty_weights meta-init fast path for NF4 prequant loads (molmoact2 / pi05 libero rSkills) — matches sim / metaworld / maniskill3 / robocasa groups
Expand All @@ -155,7 +155,7 @@ metaworld = [
"lerobot", # reason: provides lerobot.envs.metaworld used by the MetaWorld scene adapter
"gymnasium[mujoco]>=1.3.0", # reason: gym backend MetaWorldEnv is built on
"mujoco>=3.8.0", # reason: physics engine for MetaWorld scenes
"transformers>=5.4.0,<5.14.0", # reason: SmolVLA backbone (pinned to match the sim group)
"transformers>=5.4.0,<5.15.0", # reason: SmolVLA backbone (pinned to match the sim group)
"accelerate>=1.13.0", # reason: device_map + dtype handling for SmolVLA
"num2words>=0.5.14", # reason: required by SmolVLAProcessor (transformers) for number tokenization
]
Expand All @@ -167,7 +167,7 @@ maniskill3 = [
"mani-skill>=3.0.0b9",
"gymnasium>=1.0.0", # reason: gym backend ManiSkill envs are built on
"lerobot", # reason: every non-RLDX policy adapter (smolvla / pi05 / xvla / act / diffusion) imports lerobot.policies.* at load time. Pulled bare (no `[libero]`) to keep robosuite unconstrained — `mani-skill` carries SAPIEN, not robosuite, so there is no robosuite version pin to clash with. Once an MS3-specific rSkill ships in-tree this becomes load-bearing; today it future-proofs the group + matches the metaworld group's convention.
"transformers>=5.4.0,<5.14.0", # reason: SmolVLA / π0.5 / xVLA / ACT / diffusion adapters all import transformers at load time (lerobot's own dep is too loose for the Py3.12 + lerobot compat path — see python/rskill/src/openral_rskill/_lerobot_compat.py). Pinned to match the sim + metaworld groups so a `--group sim --group maniskill3` sync resolves to a single transformers version.
"transformers>=5.4.0,<5.15.0", # reason: SmolVLA / π0.5 / xVLA / ACT / diffusion adapters all import transformers at load time (lerobot's own dep is too loose for the Py3.12 + lerobot compat path — see python/rskill/src/openral_rskill/_lerobot_compat.py). Pinned to match the sim + metaworld groups so a `--group sim --group maniskill3` sync resolves to a single transformers version.
"accelerate>=1.13.0", # reason: device_map + dtype handling for SmolVLA / pi05 (matches metaworld group convention)
"num2words>=0.5.14", # reason: required by SmolVLAProcessor (transformers) for number tokenization (matches metaworld group)
]
Expand Down Expand Up @@ -268,7 +268,7 @@ robocasa = [
# pi05-vision-nf4 scene (and any future transformers-backed adapter on a
# robosuite>=1.5 task).
"lerobot", # reason: policy adapters (smolvla / pi05 / xvla / act / diffusion) import lerobot.policies.* at load time. Pulled bare (no `[libero]`) — robosuite>=1.5 above is the RoboCasa pin and conflicts with libero's robosuite==1.4.
"transformers>=5.4.0,<5.14.0", # reason: SmolVLA / π0.5 / xVLA backbones live in transformers (PaliGemma, Florence-2, SmolVLM). Pinned to match the sim / metaworld / maniskill3 groups so multi-group syncs resolve to a single version.
"transformers>=5.4.0,<5.15.0", # reason: SmolVLA / π0.5 / xVLA backbones live in transformers (PaliGemma, Florence-2, SmolVLM). Pinned to match the sim / metaworld / maniskill3 groups so multi-group syncs resolve to a single version.
"accelerate>=1.13.0", # reason: device_map + dtype handling for SmolVLA / pi05 init_empty_weights fast meta-init path.
"bitsandbytes>=0.45", # reason: 4-bit (NF4) quantization for π0.5 / openarm-vision-nf4 — the pi05 adapter raises ROSConfigError without bnb on `quantization.dtype: nf4` manifests.
"num2words>=0.5.14", # reason: required by SmolVLAProcessor (transformers) for number tokenization.
Expand Down Expand Up @@ -366,7 +366,7 @@ omdet = [
# backend (openral_runner.backends.gstreamer.omdet_turbo_detector) lazily
# imports torch + transformers; this group provides them. Required by the
# `deploy sim --object-detector-manifest <…omdet-turbo-indoor…>` leg.
"transformers>=5.4.0,<5.14.0", # reason: AutoModelForZeroShotObjectDetection (OmDet-Turbo). Pinned to match the sim / vla groups so a multi-group sync resolves to one transformers version.
"transformers>=5.4.0,<5.15.0", # reason: AutoModelForZeroShotObjectDetection (OmDet-Turbo). Pinned to match the sim / vla groups so a multi-group sync resolves to one transformers version.
"torch>=2.2", # reason: OmDet-Turbo forward pass + .to(device); CUDA when available, CPU fallback for GPU-less hosts / deploy sim.
"pillow>=10.0", # reason: BGR frame -> RGB PIL.Image for the processor.
"timm>=1.0", # reason: OmDet-Turbo's Swin backbone loads via TimmBackbone under transformers 5.3; also used to rebuild the backbone at the processor's 640² input (the pretrained checkpoint is pinned to 224² — see _fit_timm_backbone_to_processor).
Expand Down
Loading