opake-engine
Torch substrate for the Opake library. Ships everything that uses PyTorch
that downstream stacks (opake-dpsgd, opake-dpftrl,
opake-auditing, opake-patches, opake-transformers) need:
opake.api.engine.types—ClippedPytree,NoisedPytree,PerGroup,ClipState,NoiseState, paired second-moment outputs.opake.api.engine.pytree— torch-pytree ops (tree_map,partition,merge,global_norm, …).opake.api.engine.random—RngKey(uint32 tensor),key,split,fold_in,torch.Generatorhelpers.opake.api.engine.serialization—torch.Tensor/numpy.ndarrayhandler registration with the base-side serialization registry.opake.api.engine.distributed— DDP collectives, sync registry, detection helpers.opake.api.engine.noise_allocation— per-group / paired-stream noise stddev math, shared between DP-SGD and DP-FTRL.opake.api.engine.clipping— fixed + AUTO-S clipping primitives (constant-sensitivity; usable by both DP-SGD and DP-FTRL).opake.api.engine.functional—make_functional,with_batch_dim,empty_collate.opake.api.engine.scheduling— step-indexed schedules + warmup / restarts composition.opake.api.engine.profiling— memory + step timer.
User-facing façades live at opake.types, opake.pytree,
opake.random, opake.distributed, opake.functional,
opake.scheduling, opake.profiling. No opake.clipping
façade — clipping is reached via stack façades
(opake.dpsgd.clipping, opake.dpftrl.clipping).
1 maintainer
Metadata
1 maintainer
Release files for opake-engine 0.16.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| opake_engine-0.16.0-py3-none-any.whl | Python 3 | none | any | Details |
Release files / opake_engine-0.16.0-py3-none-any.whl
| Download URL | opake_engine-0.16.0-py3-none-any.whl |
|---|---|
| Size | 117.8 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
dfbce80857cae78a67fce9bd8e772412c1099ba1e8c7aa80dfda2573120fe6e6
|
|
BLAKE2b-256 checksum How to use checksums |
bb77ecb20334f950c3c215c331a7b39ae67893a89fd45bad0400485a5986129a
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Sep 23, 2026.
Transparency log
