Merge pull request #1016 from YosefLab/michael/tonumpy

changes numpy.array() on tensors to .numpy()
scverse · Apr 6, 2021 · f3a1080 · f3a1080
2 parents 3e4d2ff + 966dcc4
commit f3a1080
Show file tree

Hide file tree

Showing 5 changed files with 9 additions and 9 deletions.
diff --git a/scvi/external/cellassign/_model.py b/scvi/external/cellassign/_model.py
@@ -113,7 +113,7 @@ def predict(self) -> pd.DataFrame:
             gamma = outputs["gamma"]
             predictions += [gamma.cpu()]
         return pd.DataFrame(
-            np.array(torch.cat(predictions)), columns=self.cell_type_markers.columns
+            torch.cat(predictions).numpy(), columns=self.cell_type_markers.columns
         )
 
     def train(

diff --git a/scvi/model/_scanvi.py b/scvi/model/_scanvi.py
@@ -279,7 +279,7 @@ def predict(
                 pred = pred.argmax(dim=1)
             y_pred.append(pred.detach().cpu())
 
-        y_pred = np.array(torch.cat(y_pred))
+        y_pred = torch.cat(y_pred).numpy()
         if not soft:
             predictions = []
             for p in y_pred:

diff --git a/scvi/model/_totalvi.py b/scvi/model/_totalvi.py
@@ -301,7 +301,7 @@ def get_latent_library_size(
             else:
                 library = outputs["library_gene"]
             libraries += [library.cpu()]
-        return np.array(torch.cat(libraries))
+        return torch.cat(libraries).numpy()
 
     @torch.no_grad()
     def get_normalized_expression(
@@ -1025,7 +1025,7 @@ def get_protein_background_mean(self, adata, indices, batch_size):
         for tensors in scdl:
             _, inference_outputs, _ = self.module.forward(tensors)
             b_mean = inference_outputs["py_"]["rate_back"]
-            background_mean += [np.array(b_mean.cpu())]
+            background_mean += [b_mean.cpu().numpy()]
         return np.concatenate(background_mean)
 
 

diff --git a/scvi/model/base/_rnamixin.py b/scvi/model/base/_rnamixin.py
@@ -483,13 +483,13 @@ def get_likelihood_parameters(
 
             n_batch = px_rate.size(0) if n_samples == 1 else px_rate.size(1)
 
-            px_r = np.array(px_r.cpu())
+            px_r = px_r.cpu().numpy()
             if len(px_r.shape) == 1:
                 dispersion_list += [np.repeat(px_r[np.newaxis, :], n_batch, axis=0)]
             else:
                 dispersion_list += [px_r]
-            mean_list += [np.array(px_rate.cpu())]
-            dropout_list += [np.array(px_dropout.cpu())]
+            mean_list += [px_rate.cpu().numpy()]
+            dropout_list += [px_dropout.cpu().numpy()]
 
         dropout = np.concatenate(dropout_list)
         means = np.concatenate(mean_list)
@@ -553,4 +553,4 @@ def get_latent_library_size(
             else:
                 library = torch.distributions.LogNormal(ql_m, ql_v.sqrt()).mean
             libraries += [library.cpu()]
-        return np.array(torch.cat(libraries))
+        return torch.cat(libraries).numpy()
diff --git a/scvi/model/base/_vaemixin.py b/scvi/model/base/_vaemixin.py
@@ -177,4 +177,4 @@ def get_latent_representation(
                     z = qz_m
 
             latent += [z.cpu()]
-        return np.array(torch.cat(latent))
+        return torch.cat(latent).numpy()