Specifically, we should (1) make it automated so we can run all the tests easily, (2) add a few of the competitive examples cited here, (3) document things, (4) try to expand the coverage of toolkits to at least PyTorch, maybe others: https://github.com/clab/dynet/issues/1284
Currently we have the DyNet benchmarks repository that needs to be updated:
Specifically, we should (1) make it automated so we can run all the tests easily, (2) add a few of the competitive examples cited here, (3) document things, (4) try to expand the coverage of toolkits to at least PyTorch, maybe others: https://github.com/clab/dynet/issues/1284