Skip to main content

mbirtorch

CI Documentation Status

A PyTorch port of mbirjax: high-performance tomographic reconstruction with multi-GPU support.

This repo is the home for the port evaluated in mbirjax_plans/plans/torch_port/port_plan.md (checked out parallel to this repo). The plan defines the motivation, the parity gates against mbirjax, the incremental phase plan, and a progress record; its findings pages and the supporting scripts live alongside it under mbirjax_plans/plans/.

Setup

Create the conda environment and install the package in editable mode:

conda env create -f environment.yml
conda activate mbirtorch
pip install -e ".[test]"

Tests

dev_scripts/run_tests.sh

Caches

mbirtorch keeps one on-disk cache: compiled torch.compile artifacts, under ~/.mbirtorch/torch_cache. It exists to make cold starts fast -- with it, a fresh process reuses prior compilations instead of recompiling (roughly 14 s down to 2 s for a first small reconstruction). It grows with the number of distinct compiled shapes and typically stays in the tens of megabytes; it is never cleaned automatically. To remove it:

import mbirtorch
mbirtorch.clear_cache()   # deletes ~/.mbirtorch entirely (recreated empty)

The location can be redirected by setting the TORCHINDUCTOR_CACHE_DIR environment variable before the first compile (e.g. to node-local or scratch storage on a cluster, where home quotas are tight); clear_cache() does not touch a redirected location.

Everything else the package caches is in-memory only and is freed with the objects that hold it (e.g. the per-model pixel-index cache); nothing besides ~/.mbirtorch is written to disk.

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

mbirtorch-0.0.1.tar.gz (647.4 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

mbirtorch-0.0.1-py3-none-any.whl (537.3 kB view details)

Uploaded Python 3

File details

Details for the file mbirtorch-0.0.1.tar.gz.

File metadata

  • Download URL: mbirtorch-0.0.1.tar.gz
  • Upload date:
  • Size: 647.4 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/7.0.0 CPython/3.13.14

File hashes

Hashes for mbirtorch-0.0.1.tar.gz
Algorithm Hash digest
SHA256 5bd558c6083e6dd1686e7c0613687b7f3a2d92f669b3f8b2cb9be0f8b7a7401e
MD5 f8e735af1bae9e74c94146f41e9057c0
BLAKE2b-256 0019fb97590b82aef21e91fd94b7f38619912a801002b9deb4952ec8dcc5b170

See more details on using hashes here.

Provenance

The following attestation bundles were made for mbirtorch-0.0.1.tar.gz:

Publisher: release.yml on cabouman/mbirtorch

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file mbirtorch-0.0.1-py3-none-any.whl.

File metadata

  • Download URL: mbirtorch-0.0.1-py3-none-any.whl
  • Upload date:
  • Size: 537.3 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/7.0.0 CPython/3.13.14

File hashes

Hashes for mbirtorch-0.0.1-py3-none-any.whl
Algorithm Hash digest
SHA256 515a16414478b2a21eef275c610169aeee7a76abe53859c0b8ac3353cabc3ffc
MD5 ac2530a88a85c823da2dec1906f8ee68
BLAKE2b-256 e377ba0b2ae0021443e4f950be85ea612c0edb64365149c907c6fbf42558593c

See more details on using hashes here.

Provenance

The following attestation bundles were made for mbirtorch-0.0.1-py3-none-any.whl:

Publisher: release.yml on cabouman/mbirtorch

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Release history Release notifications | RSS feed

0.0.2

2 files

This release

0.0.1 This release

2 files

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page