This repository shows how to use Adaptive Implicit MLE (links to our NeurIPS 2021, AAAI 2023, videos, slides, etc. are available here: https://github.com/EdinburghNLP/torch-adaptive-imle/) for implementing a differentiable ranking function in 10 minutes and, for example, train a model to directly optimise the training Mean Reciprocal Rank.
It is based on the torch-adaptive-imle library:
$ pip install -U git+https://github.com/EdinburghNLP/torch-adaptive-imle.git