Commit Graph
4 Commits
Author SHA1 Message Date
Zhengxing Chen e3fcbb639e Support padding in Seq2Reward
Summary:
Coordinate data fetcher and batch preprocessor to add padding, valid_seq_len, and valid_next_seq_len.

Use padding information in Seq2Reward training

Reviewed By: kittipatv, xuruiyang

Differential Revision: D24239068

fbshipit-source-id: 26028b6a55695023694caa3ff87aae4d2459266f
2020-10-13 21:52:36 -07:00
Badri Narayan Bhaskar 8fe9badf22 SlateScoreBatchProcessor
Summary: As titled

Reviewed By: kaiwenw

Differential Revision: D23549686

fbshipit-source-id: 56eca7e4126a46ab7f592218ead92c3f7bd64900
2020-09-07 15:20:17 -07:00
Ruiyang Xu 49e4d26415 WM DQN Eval
Summary: World Model DQN Eval Ready for Product

Reviewed By: czxttkl

Differential Revision: D22450804

fbshipit-source-id: c61ee9c27e253e27231cbdc7848ffe987c740cb2
2020-07-19 22:58:21 -07:00
Kaiwen Wang c5de55e7b9 Update DiscreteDQNBatchPreprocessor to use transforms
Reviewed By: czxttkl

Differential Revision: D21920081

fbshipit-source-id: 6c6f109898d5443cd1a420983cef153da7384b9b
2020-06-10 16:09:25 -07:00