-
Notifications
You must be signed in to change notification settings - Fork 12
Add files from Turing repo #12
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.
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
Conversation
- Dict --> Array - push! : preallocate memory - forkc: remove forkc for the particles with only one chilcren.
Appveyor test
# Conflicts: # README.md
Benchmarks
Moved some untested functions to a separate branch.
Turing now works with Julia 0.5
# Conflicts: # README.md # docs/source/index.rst
# Conflicts: # src/core/compiler.jl # test/assume.jl # test/clonetask.jl # test/noparam.jl # test/predict.jl # test/runtests.jl
New interface and Gibbs sampling
Unify Mamba.Chain and Turing.Chain.
Unify rand interface
# Conflicts: # src/samplers/hmc.jl # src/samplers/sampler.jl # test/gibbs.jl
Remove sample macro
Refactor VarInfo
* Update Turing to AbstractMCMC 0.4 * Update sampling interface guide * Simplify resume_from handling in particle methods * Actually pass the RNG * Use cat instead of vcat to avoid mysterious type issues * Revert hack * Add error message Co-authored-by: Hong Ge <[email protected]>
* squash commit * minor fixes * bump DAD compat version * test fixes * use ForwardDiff for non-AD tests * Update Project.toml Co-Authored-By: David Widmann <[email protected]> * apply David's suggestions * Update src/Turing.jl Co-Authored-By: David Widmann <[email protected]> * David's comment * David's and Will's comments * Update Project.toml Co-Authored-By: David Widmann <[email protected]> * Update Project.toml Co-Authored-By: David Widmann <[email protected]> * Update src/core/ad.jl Co-Authored-By: David Widmann <[email protected]> * Update src/core/ad.jl Co-Authored-By: David Widmann <[email protected]> * Update src/core/ad.jl Co-Authored-By: David Widmann <[email protected]> * indentation Co-authored-by: David Widmann <[email protected]>
* Remove distribution wrappers (~> DynamicPPL) * Remove asssume/observer fallback code (~> DynamicPPL) * Update imports/exports/qualified names * Remove unnecessary code from utils (is in DPPL now) * fix rebase * bump DPPL compat version Co-authored-by: mohamed82008 <[email protected]>
Update Turing to AbstractMCMC 1.0, AdvancedMH 0.5, and DynamicPPL 0.6
* Use `propagate!` instead of `Libtask.consume` * Fix tests
* Make Turing compatible with DynamicPPL 0.7 * Uncomment test that should be fixed upstream * Updating variables in MH currently only works for `VarInfo` * Use DynamicPPL 0.7.1
* Allow sampling from the prior * Apply suggestions from code review Co-authored-by: Hong Ge <[email protected]> * Use `getlogp` instead of `getlp` * Add seed and increase number of samples Co-authored-by: Hong Ge <[email protected]>
* Fix particle filters with adaptive resampling and add documentation Fixes TuringLang/DynamicPPL.jl#104 * Fix and extend tests of `ParticleContainer` * Move logevidence tests from DynamicPPL * Add more convenient constructors for Particle Gibbs * Relax type annotations * Check for approximate equality only * Add docstring and reference
* Make GibbsComponent a simple trait (fix 1306) * Simplify isgibbscomponent; implement some other suggestions * Temporary fix for Distributions dependency
Co-authored-by: David Widmann <[email protected]>
* Remove `Turing.DEBUG` * Remove `_debug` * Update advi.jl Co-authored-by: Hong Ge <[email protected]>
Co-authored-by: Cameron Pfiffer <[email protected]>
Thanks, @devmotion - this is very helpful! |
@devmotion maybe we can rename these files in a new PR, so that changes in #11 can be recognised by Github? |
Sure, I just wanted to keep this PR clean. |
Maybe it would be good as well to generate the default package layout with an empty module in a separate PR first, e.g., with PkgTemplates? |
Sounds good to me! |
This PR adds relevant files from the Turing repo in a
Turing/
subfolder, together with their history.The files were extracted with git-filter-repo:
Hopefully this simplifies reviews of PRs such as #11 (see, e.g., #11 (comment)) since it would be clearer what parts of the existing implementation in Turing were changed. Moreover, by keeping the history it might be easier to track down changes and understand the implementation better.