I am exploring using this for training TransE embeddings had some beginner questions:
Have been using Pytorch BigGraph , single box mode - For TransE did anybody do a benchmarking of quality on standard datasets , what are the key differences between the two frameworks ?
I want to run incremental training, ie; once I have trained embeddings and graph gets updated, only retrain for the new nodes to avoid volatility of existing embeddings. Is there a way to do that here?
Can I run this for a graph with 100 million edges , that fit into a single box ?
I am starting dig into the docs to understand the framework better, so apologies for if some these questions are already covered in the docs.
I am exploring using this for training TransE embeddings had some beginner questions:
I am starting dig into the docs to understand the framework better, so apologies for if some these questions are already covered in the docs.