How to convert SparseVectors inside tabular data to a Tensor

I have have a tabular data set that I am wanting to use for my features. It consists of numerical values and categorical variables that I have one hot encoded.

I want to but these features into a tensor but when i convert to a numpy array then try to place into a tensor it says an error due to the sparse vectors.

Array data:

Is there another way to convert categorical variables instead of one hot encoding? I have seen people mention about nn.embeddings but not sure how that would work if my case if I am feeding in features with some numerical and some categorical