Hi there, I wanted to ask if the trained weights of the model used and evaluated in the paper are available somewhere? It would be nice to be able to reproduce the work in the paper without having to re-train. Plus, to my understanding, even with the same training script and data I won't be able to get the exact same results without an identical initialization and hardware setup. I noticed the comment here in an earlier issue, but just wanted to double-check. Apologies if the weights are already available and I'm just missing them.
Hi there, I wanted to ask if the trained weights of the model used and evaluated in the paper are available somewhere? It would be nice to be able to reproduce the work in the paper without having to re-train. Plus, to my understanding, even with the same training script and data I won't be able to get the exact same results without an identical initialization and hardware setup. I noticed the comment here in an earlier issue, but just wanted to double-check. Apologies if the weights are already available and I'm just missing them.