TypeError: can't convert cuda:0 device type tensor to numpy. Use Tensor.cpu() to copy the tensor to host memory first.

[chuangzhao0601]

Hello author, when I reproduce the DeepTalk-singleST.ipynb file, I encountered such error:
dt.File_Train(st_data, pathways, lrpairs_train, meta_data, species, LR_train = 'Sst_Sstr2',
outdir = Test_dir)

dt.data_for_train(st_data,data_dir='/mnt/workspace/test/single-cell/data/',LR_train="Sst_Sstr2")

dt.Train(data_name='Sst_Sstr2',
data_path='/mnt/workspace/test/single-cell/data/',
outdir='/mnt/workspace/test/single-cell/data/Sst_Sstr2/output',
pretrained_embeddings='/mnt/workspace/test/single-cell/data/Sst_Sstr2/data_pca.emd',
n_epochs=30, ft_n_epochs=10)

Loading graph...
DataSet Stats :
Number of Nodes 2399
Number of Edges 119477
Number of Node types 1
Number of relations 1
Graph with 2399 nodes and 119477 edges
2399 2
check torch.Size([2399, 128]) 2399
node2vec tensor torch.Size([2399, 128])
No. of nodes with pretrained embedding: 2399
No. edges in test data: 48252
PRETRAINING

generate walks ...
no. of walks 19192
train /mnt/workspace/test/single-cell/data//Sst_Sstr2/outdir/Sst_Sstr2_walks_train.txt
validate /mnt/workspace/test/single-cell/data//Sst_Sstr2/outdir/Sst_Sstr2_walks_validate.txt
test /mnt/workspace/test/single-cell/data//Sst_Sstr2/outdir/Sst_Sstr2_walks_test.txt

Run model for pre-training ...

Begin Training

Epoch: 0
MinLoss: 1000000.0, CurLoss: 7.9438
epoch time: (s) 138.76754474639893

Epoch: 10
MinLoss: 1000000.0, CurLoss: 7.8022
epoch time: (s) 138.13774728775024

Epoch: 20
MinLoss: 1000000.0, CurLoss: 7.7198
epoch time: (s) 140.38009071350098

Best Epoch: 29
In GenericGraph link Prediction Generation
168882
Generating false edges by random false target
train
valid
test

!!!!
4 4 Linear(in_features=512, out_features=512, bias=True)
Evaluating valid batch 0
Evaluating valid batch 100
Evaluating test batch 0
Evaluating test batch 100
Evaluating test batch 200

Begin evaluation for link prediction...
TypeError: can't convert cuda:0 device type tensor to numpy. Use Tensor.cpu() to copy the tensor to host memory first.

the important package versions:
deeptalk-st 0.0.4
as for other packages,I followed the Installation guidance