Info: Zenodo’s user support line is staffed on regular business days between Dec 23 and Jan 5. Response times may be slightly longer than normal.

Published October 22, 2020 | Version v1
Dataset Open

Dataset and model weights for paper "Multi-Referenced Training for Dialogue Response Generation"

Creators

  • 1. Kyoto University

Description

dataset.txt: JSON file of dataset which has multiple references in each training sample

gpt2_medium.floor_rel.seed_42.20200407-135521.model.pt: model weights of the finetuned GPT-2 used as a seq-level teacher model

gpt2_small.floor_none.seed_42.20200407-133531.model.pt: model weights of the finetuned GPT-2 used as a token-level teacher model (because medium GPT-2 is too large and too slow for token-level KD)

roberta_large.floor_none.seed_42.2020-04-01-12_28_50.semi.supervised_by_overall.model.pt: model weights of Roberta-eval for evaluating

mturk_results.json: JSON file of Amazon MTurk human evaluation results

Files

dataset.txt

Files (4.5 GB)

Name Size Download all
md5:b5844be7c076a2f5a6f23ac829fdb0e6
1.0 GB Preview Download
md5:ca2f354dd68da70f06500baab0c0c025
1.5 GB Download
md5:0524ed7bc7aa7dd4fcc840e77055d746
548.1 MB Download
md5:5ea91104d1735574033ed9b13d031465
872.9 kB Preview Download
md5:f5fee852e5e8fdc69647480d76ec6ff5
1.4 GB Download