-
Notifications
You must be signed in to change notification settings - Fork 341
/
_jaxscvi.py
165 lines (139 loc) · 4.96 KB
/
_jaxscvi.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
import logging
from collections.abc import Sequence
from typing import Literal, Optional
import jax.numpy as jnp
import numpy as np
from anndata import AnnData
from scvi import REGISTRY_KEYS
from scvi.data import AnnDataManager
from scvi.data.fields import CategoricalObsField, LayerField
from scvi.module import JaxVAE
from scvi.utils import setup_anndata_dsp
from .base import BaseModelClass, JaxTrainingMixin
logger = logging.getLogger(__name__)
class JaxSCVI(JaxTrainingMixin, BaseModelClass):
"""``EXPERIMENTAL`` single-cell Variational Inference :cite:p:`Lopez18`, but with a Jax backend.
This implementation is in a very experimental state. API is completely subject to change.
Parameters
----------
adata
AnnData object that has been registered via :meth:`~scvi.model.JaxSCVI.setup_anndata`.
n_hidden
Number of nodes per hidden layer.
n_latent
Dimensionality of the latent space.
dropout_rate
Dropout rate for neural networks.
gene_likelihood
One of:
* ``'nb'`` - Negative binomial distribution
* ``'poisson'`` - Poisson distribution
**model_kwargs
Keyword args for :class:`~scvi.module.JaxVAE`
Examples
--------
>>> adata = anndata.read_h5ad(path_to_anndata)
>>> scvi.model.JaxSCVI.setup_anndata(adata, batch_key="batch")
>>> vae = scvi.model.JaxSCVI(adata)
>>> vae.train()
>>> adata.obsm["X_scVI"] = vae.get_latent_representation()
"""
_module_cls = JaxVAE
def __init__(
self,
adata: AnnData,
n_hidden: int = 128,
n_latent: int = 10,
dropout_rate: float = 0.1,
gene_likelihood: Literal["nb", "poisson"] = "nb",
**model_kwargs,
):
super().__init__(adata)
n_batch = self.summary_stats.n_batch
self.module = self._module_cls(
n_input=self.summary_stats.n_vars,
n_batch=n_batch,
n_hidden=n_hidden,
n_latent=n_latent,
dropout_rate=dropout_rate,
gene_likelihood=gene_likelihood,
**model_kwargs,
)
self._model_summary_string = ""
self.init_params_ = self._get_init_params(locals())
@classmethod
@setup_anndata_dsp.dedent
def setup_anndata(
cls,
adata: AnnData,
layer: Optional[str] = None,
batch_key: Optional[str] = None,
**kwargs,
):
"""%(summary)s.
Parameters
----------
%(param_adata)s
%(param_layer)s
%(param_batch_key)s
"""
setup_method_args = cls._get_setup_method_args(**locals())
anndata_fields = [
LayerField(REGISTRY_KEYS.X_KEY, layer, is_count_data=True),
CategoricalObsField(REGISTRY_KEYS.BATCH_KEY, batch_key),
]
adata_manager = AnnDataManager(fields=anndata_fields, setup_method_args=setup_method_args)
adata_manager.register_fields(adata, **kwargs)
cls.register_manager(adata_manager)
def get_latent_representation(
self,
adata: Optional[AnnData] = None,
indices: Optional[Sequence[int]] = None,
give_mean: bool = True,
n_samples: int = 1,
batch_size: Optional[int] = None,
) -> np.ndarray:
r"""Return the latent representation for each cell.
This is denoted as :math:`z_n` in our manuscripts.
Parameters
----------
adata
AnnData object with equivalent structure to initial AnnData. If `None`, defaults to the
AnnData object used to initialize the model.
indices
Indices of cells in adata to use. If `None`, all cells are used.
give_mean
Whether to return the mean of the posterior distribution or a sample.
n_samples
Number of samples to use for computing the latent representation.
batch_size
Minibatch size for data loading into model. Defaults to `scvi.settings.batch_size`.
Returns
-------
latent_representation : np.ndarray
Low-dimensional representation for each cell
"""
self._check_if_trained(warn=False)
adata = self._validate_anndata(adata)
scdl = self._make_data_loader(
adata=adata, indices=indices, batch_size=batch_size, iter_ndarray=True
)
jit_inference_fn = self.module.get_jit_inference_fn(
inference_kwargs={"n_samples": n_samples}
)
latent = []
for array_dict in scdl:
out = jit_inference_fn(self.module.rngs, array_dict)
if give_mean:
z = out["qz"].mean
else:
z = out["z"]
latent.append(z)
concat_axis = 0 if ((n_samples == 1) or give_mean) else 1
latent = jnp.concatenate(latent, axis=concat_axis)
return self.module.as_numpy_array(latent)
def to_device(self, device):
pass
@property
def device(self):
return self.module.device