pandas version conflict

[GatewayPhd]

scarches_model = sca.models.SCVI.load_query_data(
    adata=adata_to_map_augmented,
    reference_model="./reference_model",
    freeze_dropout=True,
)

"...ModuleNotFoundError: No module named 'pandas.core.indexes.numeric'"

my pandas version: 2.2.2

It seems that in the lastest version of pandas, the module pandas.core.indexes.numeric does not exist. Should I reinstall and change to an older version? Or is there any other solution？Thanks.

[Koncopd]

Hi, could you post the full traceback?

[GatewayPhd]

Hi, could you post the full traceback?

Sure, here is the full traceback

INFO File ./reference_model/model.pt already downloaded

ModuleNotFoundError Traceback (most recent call last)
/tmp/ipython-input-1727706822.py in <cell line: 0>()
----> 1 scarches_model = sca.models.SCVI.load_query_data(
2 adata=adata_to_map_augmented,
3 reference_model="./reference_model",
4 freeze_dropout=True,
5 )

5 frames
/usr/local/lib/python3.12/dist-packages/scvi/model/base/_archesmixin.py in load_query_data(cls, adata, reference_model, registry, inplace_subset_query_vars, accelerator, device, unfrozen, freeze_dropout, freeze_expression, freeze_decoder_first_layer, freeze_batchnorm_encoder, freeze_batchnorm_decoder, freeze_classifier, transfer_batch, datamodule)
113 )
114
--> 115 attr_dict, var_names, load_state_dict, pyro_param_store = _get_loaded_data(
116 reference_model, device=device, adata=adata
117 )

/usr/local/lib/python3.12/dist-packages/scvi/model/base/_archesmixin.py in _get_loaded_data(reference_model, device, adata)
471 def _get_loaded_data(reference_model, device=None, adata=None):
472 if isinstance(reference_model, str):
--> 473 attr_dict, var_names, load_state_dict, _ = _load_saved_files(
474 reference_model, load_adata=False, map_location=device
475 )

/usr/local/lib/python3.12/dist-packages/scvi/model/base/_save_load.py in _load_saved_files(dir_path, load_adata, prefix, map_location, backup_url)
81 try:
82 _download(backup_url, dir_path, model_file_name)
---> 83 model = torch.load(model_path, map_location=map_location, weights_only=False)
84 except FileNotFoundError as exc:
85 raise ValueError(

/usr/local/lib/python3.12/dist-packages/torch/serialization.py in load(f, map_location, pickle_module, weights_only, mmap, **pickle_load_args)
1528 except pickle.UnpicklingError as e:
1529 raise pickle.UnpicklingError(_get_wo_message(str(e))) from None
-> 1530 return _load(
1531 opened_zipfile,
1532 map_location,

/usr/local/lib/python3.12/dist-packages/torch/serialization.py in _load(zip_file, map_location, pickle_module, pickle_file, overall_storage, **pickle_load_args)
2120 global _serialization_tls
2121 _serialization_tls.map_location = map_location
-> 2122 result = unpickler.load()
2123 _serialization_tls.map_location = None
2124

/usr/local/lib/python3.12/dist-packages/torch/serialization.py in find_class(self, mod_name, name)
2109 pass
2110 mod_name = load_module_mapping.get(mod_name, mod_name)
-> 2111 return super().find_class(mod_name, name)
2112
2113 # Load the data (which may in turn use persistent_load to load tensors)

ModuleNotFoundError: No module named 'pandas.core.indexes.numeric'

---

NOTE: If your import is failing due to a missing package, you can
manually install dependencies using either !pip or !apt.

To view examples of installing some common dependencies, click the
"Open Examples" button below.

Thanks!

[LeslieDsr]

In case this is of any help, same issue found on scvi-tools, and has been reported to be on PyTorch side : scverse/scvi-tools#2387 (comment)
(Also, many thanks to the scArches team for the great tool.)
Good day!