Bug using "scvi-tools/tabula-sapiens-heart-stereoscope" hugging face scvi model - scvi-tools 1.0.4

[LysSanzMoreta]

Hi!

I followed this tutorial scvi-hugging-face to load pre-trained scvi models and make some predictions/obtain the latent representation.

This is my current environment:

env.yaml

name: base channels:

• conda-forge
• pytorch
• defaults
• nvidia
• bioconda dependencies:
• pyopenssl
• python=3.9
• requests
• pip:
  • -r requirements.txt
• numpy
• pytorch::pytorch=2.1=py3.9_cuda12.1*
• torchvision=>0.16
• torchaudio
• biopython
• h5py
• scipy
• scikit-learn
• pandas
• matplotlib
• hdf5plugin
• pyro-ppl
• scanpy
• dill
• umap-learn
• seaborn
• dataframe_image
• xmltodict
• nbconvert
• bioconda::cellxgene
• conda-forge::lightning
• numba
• conda-forge::scvi-tools==1.0.4
• conda-forge::jax[cuda12_pip]
• conda-forge::jaxlib
• conda-forge::huggingface_hub
• conda-forge::leidenalg

requirements.txt

gget cellxgene-census

I decided to start with this example:

from scvi.hub import HubModel
torch.set_float32_matmul_precision('high')
scvi.settings.seed = 0
repository = "scvi-tools/tabula-sapiens-heart-stereoscope"
hubmodel = HubModel.pull_from_huggingface_hub(repository, cache_dir=None, revision="main")
print(hubmodel.model)

However, there was some issue with loading the model due to some attributes inconsistency.

Fetching 3 files: 100%|██████████| 3/3 [00:00<00:00, 11904.36it/s]
INFO     Loading model...                                                       
Traceback (most recent call last):
  File ".../test_scanvi.py", line 163, in <module>
    print(hubmodel.model)
  File "/home/.../micromamba/lib/python3.9/site-packages/scvi/hub/hub_model.py", line 230, in model
    self.load_model()
  File "/home/.../micromamba/lib/python3.9/site-packages/scvi/hub/hub_model.py", line 275, in load_model
    model_cls = getattr(python_module, model_cls_name)
AttributeError: module 'scvi.model' has no attribute 'RNAStereoscope'

I have been able to run this one:

from scvi.hub import HubModel
torch.set_float32_matmul_precision('high')
scvi.settings.seed = 0
repository = "scvi-tools/heart-cell-atlas-scvi"
hubmodel = HubModel.pull_from_huggingface_hub(repository, cache_dir=None, revision="main")
print(hubmodel.model)

Is this error expected due to an error in my implementation or the hugging face interface is still experimental and some errors are expected?

Thank you in advance for your time!

[martinkim0]

This indeed an issue on our end, apologies. I'll get this fixed ASAP!