slinderman
commented
11 months ago Amazing! Thanks for sharing your experience.
FYI, we do have another version of recurrent SLDS in https://github.com/lindermanlab/ssm. It uses variational inference instead of PG augmentation and Gibbs sampling, but it should be a little easier to install. One day I hope to write a JAX version!
Best, Scott
hi,
For anyone trying to get rSLDS to work, it took a long time to install everything as these packages are not maintained. Here's what I did, hoping it helps someone down the line:
*starting with new conda environment from scratch:
conda create -n slds_env pip python=3.5.4 conda activate slds_env pip install numpy==1.13.3
after doing git clone for all the recommended packages: pybasicbayes, pyhsmm, pypolyagamma, pylds, pyslds, ...
cd pybasicbayes/ pip install . cd .. pip install requests conda install cython cd pyhsmm pip install . cd ../pypolyagamma/ pip install . cd ../pylds pip install . cd ../pyslds/ pip install .
git clone autoregressive from mattjj/pyhsmm-autoregressive
cd ../pyhsmm-autoregressive/
modify the setup.py doc in pyhsmm-autoregressive. The following two line need to be replaced:
eigenurl = 'http://bitbucket.org/eigen/eigen/get/3.2.6.tar.gz'
eigenurl = 'https://gitlab.com/libeigen/eigen/-/archive/3.2.6/eigen-3.2.6.tar.gz'
thedir = glob(join('deps', 'eigen-eigen-*'))[0]
thedir = glob(join('deps', 'eigen-*'))[0]
pip install . cd ../recurrent-slds/ pip install . conda install tqdm conda install scikit-learn python examples/nascar.py