scglue.models.scglue.PairedSCGLUETrainer¶
-
class
scglue.models.scglue.
PairedSCGLUETrainer
(net, lam_data=None, lam_kl=None, lam_graph=None, lam_align=None, lam_sup=None, lam_joint_cross=None, lam_real_cross=None, lam_cos=None, normalize_u=None, domain_weight=None, optim=None, lr=None, **kwargs)[source]¶ Bases:
scglue.models.scglue.SCGLUETrainer
Paired trainer for
SCGLUE
- Parameters
lam_joint_cross (
Optional
[float
]) – Joint cross-prediction weightlam_real_cross (
Optional
[float
]) – Real cross-prediction weightnormalize_u (
Optional
[bool
]) – Whether to L2 normalize cell embeddings before decoderdomain_weight (
Optional
[Mapping
[str
,float
]]) – Relative domain weight (indexed by domain name)**kwargs – Additional keyword arguments are passed to the optimizer constructor
Methods
Compute loss functions
Format data tensors
Attributes
logger