These models look great! I would love to know if there are any plans on how the wider research community could use these models from your checkpoints? Some examples in the README of running inference through a cli or just via code would be nice. Personally I would love to try them out on my usecase before committing to any training
These models look great! I would love to know if there are any plans on how the wider research community could use these models from your checkpoints? Some examples in the README of running inference through a cli or just via code would be nice. Personally I would love to try them out on my usecase before committing to any training
Would be awesome if there are any plans to integrate with your other package: https://github.com/thunlp/OpenNRE