Issues with MLP convergence in dl.multilevel_modeling #777
Description
Please make sure these conditions are met
- I have checked that this issue has not already been reported.
- I have confirmed this bug exists on the latest version of pertpy.
- (optional) I have confirmed this bug exists on the main branch.
Report
Hello, I ran the tutorial exactly with the example data. Unfortunatley I get a bunch of convergence warnings.
Code:
'''
all_results, new_mcps = dl.multilevel_modeling(
ct_subs=ct_subs,
mcp_scores=mcps,
ws_dict=ws,
confounder="gender",
)
/home/l/lkandushi/anaconda3/envs/pert_env/lib/python3.12/site-packages/statsmodels/base/model.py:607:
ConvergenceWarning: Maximum Likelihood optimization failed to converge. Check mle_retvals
warnings.warn("Maximum Likelihood optimization failed to "
/home/l/lkandushi/anaconda3/envs/pert_env/lib/python3.12/site-packages/statsmodels/regression/mixed_linear_model.py
:2200: ConvergenceWarning: Retrying MixedLM optimization with lbfgs
warnings.warn(
/home/l/lkandushi/anaconda3/envs/pert_env/lib/python3.12/site-packages/statsmodels/regression/mixed_linear_model.py
:2237: ConvergenceWarning: The MLE may be on the boundary of the parameter space.
warnings.warn(msg, ConvergenceWarning)
/home/l/lkandushi/anaconda3/envs/pert_env/lib/python3.12/site-packages/statsmodels/regression/mixed_linear_model.py
:2261: ConvergenceWarning: The Hessian matrix at the estimated parameter values is not positive definite.
warnings.warn(msg, ConvergenceWarning)
/home/l/lkandushi/anaconda3/envs/pert_env/lib/python3.12/site-packages/statsmodels/base/model.py:607:
ConvergenceWarning: Maximum Likelihood optimization failed to converge. Check mle_retvals
warnings.warn("Maximum Likelihood optimization failed to "
/home/l/lkandushi/anaconda3/envs/pert_env/lib/python3.12/site-packages/statsmodels/regression/mixed_linear_model.py
:2200: ConvergenceWarning: Retrying MixedLM optimization with lbfgs
warnings.warn(
/home/l/lkandushi/anaconda3/envs/pert_env/lib/python3.12/site-packages/statsmodels/regression/mixed_linear_model.py
:2237: ConvergenceWarning: The MLE may be on the boundary of the parameter space.
warnings.warn(msg, ConvergenceWarning)
/home/l/lkandushi/anaconda3/envs/pert_env/lib/python3.12/site-packages/statsmodels/base/model.py:607:
ConvergenceWarning: Maximum Likelihood optimization failed to converge. Check mle_retvals
warnings.warn("Maximum Likelihood optimization failed to "
/home/l/lkandushi/anaconda3/envs/pert_env/lib/python3.12/site-packages/statsmodels/regression/mixed_linear_model.py
:2200: ConvergenceWarning: Retrying MixedLM optimization with lbfgs
At some point with another dataset this went on for an hour before failing at this point:
p_val.replace(0, min(p for p in p_val if p is not None and p > 0)) in def _get_cor_zscores(self, estimate: pd.Series, p_val: pd.Series). I assume the errors are related and no p-values are produced but I wasn't able to check.
### Versions
```python
| Package | Version |
| ------- | ------- |
| pandas | 2.2.3 |
| pertpy | 0.11.2 |
| scanpy | 1.11.1 |
| anndata | 0.11.4 |
| numpy | 1.26.4 |
| Dependency | Version |
| ------------------------ | ------------------- |
| pycparser | 2.22 |
| nvidia-cuda-cupti-cu12 | 12.6.80 |
| decorator | 5.1.1 |
| umap-learn | 0.5.7 |
| nvidia-cuda-nvrtc-cu12 | 12.6.77 |
| seaborn | 0.13.2 |
| pyzmq | 26.2.0 |
| session-info2 | 0.1.2 |
| urllib3 | 2.3.0 |
| psutil | 6.1.1 |
| executing | 2.2.0 |
| lightning-utilities | 0.14.3 |
| flax | 0.10.6 |
| jupyter_client | 8.6.3 |
| toolz | 1.0.0 |
| nvidia-nvjitlink-cu12 | 12.6.85 |
| threadpoolctl | 3.6.0 |
| ipykernel | 6.29.5 |
| pillow | 11.2.1 |
| fsspec | 2025.5.0 |
| lamin_utils | 0.14.0 |
| nvidia-cusparse-cu12 | 12.5.4.2 |
| nvidia-nvtx-cu12 | 12.6.77 |
| docrep | 0.3.2 |
| filelock | 3.18.0 |
| pynndescent | 0.5.13 |
| nvidia-cufile-cu12 | 1.11.1.6 |
| msgpack | 1.1.0 |
| nvidia-nccl-cu12 | 2.26.2 |
| pyparsing | 3.2.3 |
| sparsecca | 0.3.1 |
| Pygments | 2.19.1 |
| fast-array-utils | 1.2.1 |
| parso | 0.8.4 |
| xarray | 2025.4.0 |
| scikit-learn | 1.5.2 |
| importlib_resources | 6.5.2 |
| packaging | 24.2 |
| optax | 0.2.4 |
| rich | 14.0.0 |
| patsy | 1.0.1 |
| wcwidth | 0.2.13 |
| triton | 3.3.0 |
| pytz | 2025.2 |
| sparse | 0.17.0 |
| PyYAML | 6.0.2 |
| statsmodels | 0.14.4 |
| simplejson | 3.20.1 |
| PubChemPy | 1.0.4 |
| nvidia-cufft-cu12 | 11.3.0.4 |
| asttokens | 3.0.0 |
| certifi | 2024.12.14 |
| nvidia-cudnn-cu12 | 9.5.1.17 |
| six | 1.17.0 |
| tqdm | 4.67.1 |
| nvidia-cuda-runtime-cu12 | 12.6.77 |
| typing_extensions | 4.12.2 |
| jax | 0.6.0 |
| nvidia-curand-cu12 | 10.3.7.77 |
| pyomo | 6.9.2 |
| defusedxml | 0.7.1 |
| idna | 3.10 |
| adjustText | 1.3.0 |
| ply | 3.11 |
| numba | 0.60.0 |
| ml_dtypes | 0.5.1 |
| jupyter_core | 5.7.2 |
| attrs | 25.1.0 |
| scikit-misc | 0.5.1 |
| scvi-tools | 1.3.1.post1 |
| prompt_toolkit | 3.0.50 |
| pyarrow | 20.0.0 |
| networkx | 3.4.2 |
| nvidia-cusolver-cu12 | 11.7.1.2 |
| charset-normalizer | 3.4.1 |
| ipython | 8.31.0 |
| sympy | 1.14.0 |
| jaxtyping | 0.3.2 |
| mudata | 0.3.1 |
| comm | 0.2.2 |
| legacy-api-wrap | 1.4.1 |
| cycler | 0.12.1 |
| python-dateutil | 2.9.0.post0 |
| tornado | 6.4.2 |
| lineax | 0.0.8 |
| debugpy | 1.8.12 |
| etils | 1.12.2 |
| natsort | 8.4.0 |
| h5py | 3.13.0 |
| absl-py | 2.2.2 |
| jedi | 0.19.2 |
| pyro-ppl | 1.9.1 |
| cffi | 1.17.1 |
| lightning | 2.5.1.post0 |
| opt_einsum | 3.4.0 |
| blitzgsea | 1.3.54 |
| torchmetrics | 1.7.1 |
| prompt-toolkit | 3.0.50 |
| stack-data | 0.6.3 |
| ipywidgets | 8.1.5 |
| llvmlite | 0.43.0 |
| ml_collections | 1.1.0 |
| setuptools | 75.8.0 |
| pure-eval | 0.2.3 |
| numpyro | 0.18.0 |
| kiwisolver | 1.4.8 |
| matplotlib | 3.10.3 |
| multipledispatch | 1.0.0 (0.6.0) |
| wadler_lindig | 0.1.6 |
| joblib | 1.5.0 |
| pure_eval | 0.2.3 |
| decoupler | 1.8.0 |
| traitlets | 5.14.3 |
| requests | 2.32.3 |
| matplotlib-inline | 0.1.7 |
| platformdirs | 4.3.6 |
| jaxopt | 0.8.5 |
| scipy | 1.15.3 |
| jaxlib | 0.6.0 |
| mpmath | 1.3.0 |
| torch | 2.7.0 (2.7.0+cu126) |
| ott-jax | 0.5.0 |
| equinox | 0.12.2 |
| nvidia-cublas-cu12 | 12.6.4.1 |
| pytorch-lightning | 2.5.1.post0 |
| chex | 0.1.89 |
| Component | Info |
| --------- | -------------------------------------------------------------------------------- |
| Python | 3.12.9 | packaged by Anaconda, Inc. | (main, Feb 6 2025, 18:56:27) [GCC 11.2.0] |
| OS | Linux-6.4.0-150600.23.47-default-x86_64-with-glibc2.38 |
| Updated | 2025-05-22 00:30 |