[Bug] Cannot run example data from novae:

[kenxie7]

Description

I tried to run the example code from the tutorial on the given datasets, however it didn't seem to run and encounter the following error when running compute_representation.

Code

import novae
model = novae.Novae.from_pretrained("novae-mouse-0")

model

Loading weights from local directory
Novae model
├── Known genes: 60697
├── Parameters: 32.0M
└── Model name: novae-mouse-0

# Option 1: zero-shot
adata = novae.utils.load_dataset(tissue="brain", species="mouse", pattern=".*5_7.*")[0]
model.compute_representations(adata, zero_shot=True,)

ERROR:

KeyError: 'count'

The above exception was the direct cause of the following exception:

KeyError Traceback (most recent call last)
Cell In[36], line 2
1 # Option 1: zero-shot
----> 2 model.compute_representations(adata, zero_shot=True,)# accelerator='cuda', num_workers = 20) #slide_key='sample',
4 # Option 2: fine-tuning
5 #model.fine_tune(adata)
6 #model.compute_representations(adata)

File ~/miniforge-pypy3/envs/py39_gpu_mrvi/lib/python3.9/site-packages/novae/utils/_utils.py:76, in requires_fit..wrapper(model, *args, **kwargs)
73 @wraps(f)
74 def wrapper(model, *args, **kwargs):
75 assert model.mode.trained, "Novae must be trained first, so consider running model.fit()"
---> 76 return f(model, *args, **kwargs)

File ~/miniforge-pypy3/envs/py39_gpu_mrvi/lib/python3.9/site-packages/torch/utils/_contextlib.py:115, in context_decorator..decorate_context(*args, **kwargs)
112 @functools.wraps(func)
113 def decorate_context(*args, **kwargs):
114 with ctx_factory():
--> 115 return func(*args, **kwargs)

File ~/miniforge-pypy3/envs/py39_gpu_mrvi/lib/python3.9/site-packages/novae/model.py:336, in Novae.compute_representations(self, adata, slide_key, zero_shot, accelerator, num_workers)
334 adatas = self._prepare_adatas(adata, slide_key=slide_key)
335 for adata in adatas:
--> 336 datamodule = self._init_datamodule(adata)
337 self._compute_representations_datamodule(adata, datamodule)
339 if self.mode.zero_shot:

File ~/miniforge-pypy3/envs/py39_gpu_mrvi/lib/python3.9/site-packages/novae/model.py:175, in Novae._init_datamodule(self, adata, sample_cells, **kwargs)
172 def _init_datamodule(
173 self, adata: AnnData | list[AnnData] | None = None, sample_cells: int | None = None, **kwargs: int
174 ):
--> 175 return NovaeDatamodule(
176 self._to_anndata_list(adata),
177 cell_embedder=self.cell_embedder,
178 batch_size=self.hparams.batch_size,
179 n_hops_local=self.hparams.n_hops_local,
180 n_hops_view=self.hparams.n_hops_view,
181 num_workers=self._num_workers,
182 sample_cells=sample_cells,
183 **kwargs,
184 )

File ~/miniforge-pypy3/envs/py39_gpu_mrvi/lib/python3.9/site-packages/novae/data/datamodule.py:27, in NovaeDatamodule.init(self, adatas, cell_embedder, batch_size, n_hops_local, n_hops_view, num_workers, sample_cells)
16 def init(
17 self,
18 adatas: list[AnnData],
(...)
24 sample_cells: int | None = None,
25 ) -> None:
26 super().init()
---> 27 self.dataset = NovaeDataset(
28 adatas,
29 cell_embedder=cell_embedder,
30 batch_size=batch_size,
31 n_hops_local=n_hops_local,
32 n_hops_view=n_hops_view,
33 sample_cells=sample_cells,
34 )
35 self.batch_size = batch_size
36 self.num_workers = num_workers

File ~/miniforge-pypy3/envs/py39_gpu_mrvi/lib/python3.9/site-packages/novae/data/dataset.py:69, in NovaeDataset.init(self, adatas, cell_embedder, batch_size, n_hops_local, n_hops_view, sample_cells)
66 self.single_adata = len(self.adatas) == 1
67 self.single_slide_mode = self.single_adata and len(np.unique(self.adatas[0].obs[Keys.SLIDE_ID])) == 1
---> 69 self._init_dataset()

File ~/miniforge-pypy3/envs/py39_gpu_mrvi/lib/python3.9/site-packages/novae/data/dataset.py:94, in NovaeDataset._init_dataset(self)
90 if self.single_adata:
91 self.obs_ilocs = np.array([(0, obs_index) for obs_index in self.valid_indices[0]])
93 self.slides_metadata: pd.DataFrame = pd.concat(
---> 94 [
95 self._adata_slides_metadata(adata_index, obs_indices)
96 for adata_index, obs_indices in enumerate(self.valid_indices)
97 ],
98 axis=0,
99 )
101 self.shuffle_obs_ilocs()

File ~/miniforge-pypy3/envs/py39_gpu_mrvi/lib/python3.9/site-packages/novae/data/dataset.py:95, in (.0)
90 if self.single_adata:
91 self.obs_ilocs = np.array([(0, obs_index) for obs_index in self.valid_indices[0]])
93 self.slides_metadata: pd.DataFrame = pd.concat(
94 [
---> 95 self._adata_slides_metadata(adata_index, obs_indices)
96 for adata_index, obs_indices in enumerate(self.valid_indices)
97 ],
98 axis=0,
99 )
101 self.shuffle_obs_ilocs()

File ~/miniforge-pypy3/envs/py39_gpu_mrvi/lib/python3.9/site-packages/novae/data/dataset.py:203, in NovaeDataset._adata_slides_metadata(self, adata_index, obs_indices)
201 slides_metadata = obs_counts.to_frame()
202 slides_metadata[Keys.ADATA_INDEX] = adata_index
--> 203 slides_metadata[Keys.N_BATCHES] = (slides_metadata["count"] // self.batch_size).clip(1)
204 return slides_metadata

File ~/miniforge-pypy3/envs/py39_gpu_mrvi/lib/python3.9/site-packages/pandas/core/frame.py:3807, in DataFrame.getitem(self, key)
3805 if self.columns.nlevels > 1:
3806 return self._getitem_multilevel(key)
-> 3807 indexer = self.columns.get_loc(key)
3808 if is_integer(indexer):
3809 indexer = [indexer]

File ~/miniforge-pypy3/envs/py39_gpu_mrvi/lib/python3.9/site-packages/pandas/core/indexes/base.py:3804, in Index.get_loc(self, key, method, tolerance)
3802 return self._engine.get_loc(casted_key)
3803 except KeyError as err:
-> 3804 raise KeyError(key) from err
3805 except TypeError:
3806 # If we have a listlike key, _check_indexing_error will raise
3807 # InvalidIndexError. Otherwise we fall through and re-raise
3808 # the TypeError.
3809 self._check_indexing_error(key)

KeyError: 'count'

[quentinblampey]

Hi @kenxie7, thanks for reporting!

I'm really sorry for the delay, it seems I didn't receive a notification for this issue...

According to your code, you provide only one AnnData object, yet Novae tries to be run in "multi-adata" mode, which is unexpected in that case.

Unfortunately, I was not able to reproduce the issue. Have you run other lines of code apart from this? What happens if you start a new script and run the commands below?

import novae

adata = novae.utils.load_dataset(tissue="brain", species="mouse", pattern=".*5_7.*")[0]

model = novae.Novae.from_pretrained("novae-mouse-0")

model.compute_representations(adata, zero_shot=True)