pyhacrf
pyhacrf copied to clipboard
dedupeio
:triangular_ruler: Hidden alignment conditional random field for classifying string pairs.
Adds basic `multiprocessing.Pool.map` parallelization to the model update step (one per example provided, so it's hugely parallel). This significantly increases the speed at which the models learn especially on large...
Bumps [pypa/cibuildwheel](https://github.com/pypa/cibuildwheel) from 2.16.2 to 2.16.5. Release notes Sourced from pypa/cibuildwheel's releases. v2.16.5 🐛 Fix an incompatibility with the GitHub Action and new GitHub Runner images for Windows that bundle...
![dependabot[bot] avatar](https://avatars.githubusercontent.com/in/29110?v=4 "dependabot[bot] avatar"){kind=avatar}
Bumps [docker/setup-qemu-action](https://github.com/docker/setup-qemu-action) from 2 to 3. Release notes Sourced from docker/setup-qemu-action's releases. v3.0.0 Node 20 as default runtime (requires Actions Runner v2.308.0 or later) by @crazy-max in docker/setup-qemu-action#102 Bump @actions/core...
https://github.com/dirko/pyhacrf/commit/8304161c92a800591acb0a33cd96d23e05fd06c8 Then we can get archive the pylbfgs lib.
In order to increase training speed, and therefore the size of the data we can train on we are going to move the gradient calculation to cython. - [x] vectorize...
Pretty troubling. We should investigate soon. https://travis-ci.org/dedupeio/pyhacrf/jobs/425952240
Let's investigate wehtehr we can get speed gains by dropping precision from 64 to 32 bytes.
I am working through the example in the "highered dataset" notebook, and I'm particularly interested in token-level features. But when I run this part of the code: ``` real =...
Bumps [pypa/cibuildwheel](https://github.com/pypa/cibuildwheel) from 2.19.1 to 2.19.2. Release notes Sourced from pypa/cibuildwheel's releases. Version 2.19.2 🐛 Update manylinux2014 pins to versions that support past-EoL CentOS 7 mirrors. (#1917) 🐛 Support --no-isolation...