ValueError in tabula_sapiens_tutorial.ipynb

[juanduphd]

Report

Hi,
I was trying out this notebook: https://github.com/YosefLab/popV/blob/main/docs/notebooks/tabula_sapiens_tutorial.ipynb, but I got stuck at this code block:

from popv.annotation import annotate_data
annotate_data(adata, save_path=f"{output_folder}/popv_output")

With this error:
ValueError: The number of threads must be between 1 and 2

I am using the example data: tmp_testing/LCA.h5ad

Any suggestions on how to overcome this issue?

Thanks,
Juan

Version information

No response

[canergen]

Hi, can you paste the full error message? Thanks. Can you also give it a try on Colab? It should be fine there.

[juanduphd]

Thanks for your quick reply. I am running it on the colab (https://colab.research.google.com/github/YosefLab/PopV/blob/main/docs/notebooks/tabula_sapiens_tutorial.ipynb) and I didn't change any of the code:

Here is the full error message:

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
[<ipython-input-19-4ee413451442>](https://localhost:8080/#) in <cell line: 3>()
      1 from popv.annotation import annotate_data
      2 
----> 3 annotate_data(adata, save_path=f"{output_folder}/popv_output")

11 frames
[/usr/local/lib/python3.10/dist-packages/popv/annotation.py](https://localhost:8080/#) in annotate_data(adata, methods, save_path, methods_kwargs)
     61         current_method = getattr(algorithms, method)(**methods_kwargs.pop(method, {}))
     62         current_method.compute_integration(adata)
---> 63         current_method.predict(adata)
     64         current_method.compute_embedding(adata)
     65         all_prediction_keys += [current_method.result_key]

[/usr/local/lib/python3.10/dist-packages/popv/algorithms/_celltypist.py](https://localhost:8080/#) in predict(self, adata)
     58         if adata.uns["_prediction_mode"] == "fast":
     59             self.classifier_dict["majority_voting"] = False
---> 60         predictions = celltypist.annotate(
     61             adata,
     62             model=adata.uns["_save_path_trained_models"] + "celltypist.pkl",

[/usr/local/lib/python3.10/dist-packages/celltypist/annotate.py](https://localhost:8080/#) in annotate(filename, model, transpose_input, gene_file, cell_file, mode, p_thres, majority_voting, over_clustering, use_GPU, min_prop)
     91     #over clustering
     92     if over_clustering is None:
---> 93         over_clustering = clf.over_cluster(use_GPU = use_GPU)
     94         predictions.adata = clf.adata
     95     elif isinstance(over_clustering, str):

[/usr/local/lib/python3.10/dist-packages/celltypist/classifier.py](https://localhost:8080/#) in over_cluster(self, resolution, use_GPU)
    435             logger.info("👀 Can not detect a neighborhood graph, will construct one before the over-clustering")
    436             adata = self.adata.copy()
--> 437             self.adata.obsm['X_pca'], self.adata.obsp['connectivities'], self.adata.obsp['distances'], self.adata.uns['neighbors'] = Classifier._construct_neighbor_graph(adata, use_GPU)
    438         else:
    439             logger.info("👀 Detected a neighborhood graph in the input object, will run over-clustering on the basis of it")

[/usr/local/lib/python3.10/dist-packages/celltypist/classifier.py](https://localhost:8080/#) in _construct_neighbor_graph(adata, use_GPU)
    408             fsc.pp.scale(adata, max_value=10)
    409             fsc.pp.pca(adata, n_comps=50)
--> 410         fsc.pp.neighbors(adata, n_neighbors=10, n_pcs=50)
    411         return adata.obsm['X_pca'], adata.obsp['connectivities'], adata.obsp['distances'], adata.uns['neighbors']
    412 

[/usr/local/lib/python3.10/dist-packages/scanpy/neighbors/__init__.py](https://localhost:8080/#) in neighbors(adata, n_neighbors, n_pcs, use_rep, knn, method, transformer, metric, metric_kwds, random_state, key_added, copy)
    189         adata._init_as_actual(adata.copy())
    190     neighbors = Neighbors(adata)
--> 191     neighbors.compute_neighbors(
    192         n_neighbors,
    193         n_pcs=n_pcs,

[/usr/local/lib/python3.10/dist-packages/scanpy/neighbors/__init__.py](https://localhost:8080/#) in compute_neighbors(self, n_neighbors, n_pcs, use_rep, knn, method, transformer, metric, metric_kwds, random_state)
    571         self.knn = knn
    572         X = _choose_representation(self._adata, use_rep=use_rep, n_pcs=n_pcs)
--> 573         self._distances = transformer.fit_transform(X)
    574         knn_indices, knn_distances = _get_indices_distances_from_sparse_matrix(
    575             self._distances, n_neighbors

[/usr/local/lib/python3.10/dist-packages/pynndescent/pynndescent_.py](https://localhost:8080/#) in fit_transform(self, X, y, **fit_params)
   2254             The diagonal is always explicit.
   2255         """
-> 2256         self.fit(X, compress_index=False)
   2257         result = self.transform(X=None)
   2258 

[/usr/local/lib/python3.10/dist-packages/pynndescent/pynndescent_.py](https://localhost:8080/#) in fit(self, X, compress_index)
   2172         effective_n_neighbors = self.n_neighbors + 1
   2173 
-> 2174         self.index_ = NNDescent(
   2175             X,
   2176             metric=self.metric,

[/usr/local/lib/python3.10/dist-packages/pynndescent/pynndescent_.py](https://localhost:8080/#) in __init__(self, data, metric, metric_kwds, n_neighbors, n_trees, leaf_size, pruning_degree_multiplier, diversify_prob, n_search_trees, tree_init, init_graph, init_dist, random_state, low_memory, max_candidates, max_rptree_depth, n_iters, delta, n_jobs, compressed, parallel_batch_queries, verbose)
    817         self._original_num_threads = numba.get_num_threads()
    818         if self.n_jobs != -1 and self.n_jobs is not None:
--> 819             numba.set_num_threads(self.n_jobs)
    820 
    821         if isspmatrix_csr(self._raw_data):

[/usr/local/lib/python3.10/dist-packages/numba/np/ufunc/parallel.py](https://localhost:8080/#) in set_num_threads(n)
    608     if not isinstance(n, (int, np.integer)):
    609         raise TypeError("The number of threads specified must be an integer")
--> 610     snt_check(n)
    611     _set_num_threads(n)
    612 

[/usr/local/lib/python3.10/dist-packages/numba/np/ufunc/parallel.py](https://localhost:8080/#) in snt_check(n)
    570     def snt_check(n):
    571         if n > NUMBA_NUM_THREADS or n < 1:
--> 572             raise ValueError(msg)
    573     return snt_check
    574 

ValueError: The number of threads must be between 1 and 2

[canergen]

Hi. The tutorial runs on my end without any issues. This is a call within Celltypist. The Celltypist function call actually is correct. I assume it's one instance on Colab that has issues with running it. Can you check the n_jobs argument in scanpy and otherwise check with scanpy what's going wrong here (create an issue there): https://scanpy.readthedocs.io/en/stable/generated/scanpy._settings.ScanpyConfig.n_jobs.html#scanpy._settings.ScanpyConfig.n_jobs. You can also change prediction_mode to fast, which will disable overclustering and therefore by-pass the issue.

[juanduphd]

Which hardware accelerator are you using? I am using T4 GPU (without Colab Pro). Perhaps that is the issue?

I tried prediction_mode="fast", unfortunately, it didn't resolve the issue on my side.