Thank you for sharing the code to train models for link predictions on KGs!
I am currently trying to save a trained model (vecFeatured) on the wikikgv2 dataset for link prediction by adding saved_model = torch.jit.script(model) line to the last line of the save_model function. However, when Torchscript tries to execute this line, it throws the following error:
Tried to set an attribute: job_handle on a non-class: Tensor:
smore/common/embedding/sparse_embed.py", line 66
else:
mat = self.embedding[indices]
mat.job_handle = self.dummy_job
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ <--- HERE
return mat
I was wondering whether you have any suggestion to get around this error or any suggestion in general to save a trained model from your repo to be used for inference (in a C++ library). Thanks in advance!
Hello,
Thank you for sharing the code to train models for link predictions on KGs!
I am currently trying to save a trained model (vecFeatured) on the wikikgv2 dataset for link prediction by adding
saved_model = torch.jit.script(model)
line to the last line of the save_model function. However, when Torchscript tries to execute this line, it throws the following error:I was wondering whether you have any suggestion to get around this error or any suggestion in general to save a trained model from your repo to be used for inference (in a C++ library). Thanks in advance!
Best regards, Jesun Firoz