vae.train() error

[sukyorsuki]

I have a model trained on ref_data with cell type labels

vae = sca.models.SCANVI( ref_data, "Unknown", n_layers=2, encode_covariates=True, deeply_inject_covariates=False, use_layer_norm="both", use_batch_norm="none"

use_cuda=True

)

vae.train( max_epochs= 400, use_gpu= False )

File c:\Python\lib\site-packages\scvi\model_scanvi.py:393, in SCANVI.train(self, max_epochs, n_samples_per_label, check_val_every_n_epoch, train_size, validation_size, batch_size, use_gpu, plan_kwargs, **trainer_kwargs) 382 trainer_kwargs["callbacks"] = sampler_callback 384 runner = TrainRunner( 385 self, 386 training_plan=training_plan, (...) 391 **trainer_kwargs, 392 ) --> 393 return runner()

File c:\Python\lib\site-packages\scvi\train_trainrunner.py:72, in TrainRunner.call(self) 69 if hasattr(self.data_splitter, "n_train"): 70 self.training_plan.n_obs_training = self.data_splitter.n_train ---> 72 self.trainer.fit(self.training_plan, self.data_splitter) 73 self._update_history() 75 # data splitter only gets these attrs after fit ... ---> 58 return torch.as_tensor(data) 59 elif isinstance(data, collections.abc.Mapping): 60 try:

RuntimeError: Could not infer dtype of numpy.float32

How to fix this problem? Any advise will be greatly appreciated!