RuntimeError: PyTorch is not linked with support for mps devices

[SunYong0821]

I run two spatial samples with same method and py, one successed, the other one failed with RuntimeError: PyTorch is not linked with support for mps devices

detailed:

Epoch 20000/20000: 100%|██████████| 20000/20000 [12:31:38<00:00,  2.25s/it, v_num=1, elbo_train=1.82e+7]
Sampling local variables, batch:   0%|          | 0/1 [00:00<?, ?it/s]
File "/home/ksd/bin/Python-3.10.4/lib/python3.10/site-packages/cell2location/models/_cell2location_model.py", line 338, in export_posterior
  self.samples = self.sample_posterior(**sample_kwargs)
File "/home/ksd/bin/Python-3.10.4/lib/python3.10/site-packages/scvi/model/base/_pyromixin.py", line 483, in sample_posterior
  samples = self._posterior_samples_minibatch(
File "/home/ksd/bin/Python-3.10.4/lib/python3.10/site-packages/scvi/model/base/_pyromixin.py", line 365, in _posterior_samples_minibatch
  args = [a.to(device) for a in args]
File "/home/ksd/bin/Python-3.10.4/lib/python3.10/site-packages/scvi/model/base/_pyromixin.py", line 365, in <listcomp>
  args = [a.to(device) for a in args]
RuntimeError: PyTorch is not linked with support for mps devices

my code is failed on this step:

adata_vis = mod.export_posterior(adata_vis, sample_kwargs={'num_samples': 1000, 'batch_size': mod.adata.n_obs, 'use_gpu': True})

[vitkl]

Is this error reproducible? What happens if you train the model again or simply restart the notebook and load the model?

[SunYong0821]

Is this error reproducible? What happens if you train the model again or simply restart the notebook and load the model?

no，i finally found it must be run one by one，may be CPU is not enough, but my cpu is 72C 144T

[vitkl]

What do you mean by "run one by one"? You can also try changing batch size for mod.export_posterior

[stupidstupidstupidstupid]

It 's a CPU problem. You must set 'use_gpu' as false