Tim Dettmers

To add a bit more context: I also tried to solve this on the bnb side by casting the tensor to a list before the optimizer is saved, but I...

I was thinking also about a potential other problem where the small quantization statistics are also split and communicated. This makes training slower it seems compared to 32-bit Adam. One...

Hi! Thank you, for your question! That is right, the indegree is how many outwards connections point to a node, or in other words, how many tail entities are related...

You should be able to install the library on colab. What you then do is to copy+paste the main.py file contents to colab

Hi! Thanks for raising this issue. While mathematically the logistic sigmoid should be the right thing to do, I have heard before that using a softmax actually performs better in...

ConvE uses 1-N scoring, meaning that we take the subject and relation and score it against _all_ other objects at once (subject, relation, E). This is done by multiplying by...

Yeah, that can be a bit confusing. It is a bit weird software artifact. It is 11 (normal) + 11 (reverse) + 1 OOV relation + 1 "" relation. The...

Yes, you are correct — thank you for reporting this! I will need to study if the results change only slightly or significantly. If the difference is only slightly I...

With the fixed label smoothing, one needs much higher label smoothing values to get a good score. Here the results of my grid search. The mean metric value is the...

This seems to be a Windows related problem. There is no `HOME` variable in the path that points to the folder where the data is stored. This is just a...