nutpie
nutpie copied to clipboard
pymc-devs
Python wrapper for nuts-rs
This repairs the householder transformation layers, changes the asymmetric transformation a bit and brings a couple more normalizing flow options for experimentation. @jessegrabowski Maybe you want to check if this...
Hello, Nutpie is excellent work and significantly accelerates our code via GPU. However, currently, it only leverages 1 GPU on Azure while compiling. Is there any way for the code...
The code ``` code = """ stan_model """ compiled = nutpie.compile_stan_model(code=code) compiled = compiled.with_data(mu=3.) trace = nutpie.sample(compiled, save_warmup=True, low_rank_modified_mass_matrix=True, store_mass_matrix=True) ``` raises ValueError: cannot reshape array of size 0 into...
The elemwise log-likelihood is not stored in the `InferenceData` that nutpie returns, even when asking for it. The following for instance doesn't error out, but doesn't add a `log_likelihood` group...
~The following are listed in the docstring for `sample` but do not appear to be used anywhere:~ - ~num_try_init~ - ~store_divergences~ - ~store_unconstrained~ - ~store_gradient~ - ~store_mass_matrix~ - ~target_accept~ -...
I'm running into jax backend issues when running a model that samples without error under the default backend. Switching to the jax backend via: ```python pm.sample(1000, tune=1000, chains=2, nuts_sampler='nutpie', nuts_sampler_kwargs={'backend':...
If you enable the normalizing flow adaptation, but don't have the required dependencies installed, the error only shows up once it tries to sample in one of the child processes:...
We should refactor the code so that all modules that are optional have wrapper modules that can always be imported, but are empty (or contain stubs that just error out)...
Following pymc-devs/pymc#7687 it would be great to have a `zarr` native storage backend for nutpie. The layout defined in `pymc.backends.zarr.ZarrTrace` is simple enough to copy and be used generically. `pymc`...
