datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
imdb_prefix20_forDPO_gpt2-large-imdb-FT_siebert_sentiment-roberta-large-english
Dataset Card for "imdb_prefix20_forDPO_gpt2-large-imdb-FT_siebert_sentiment-roberta-large-english"
1. Purpose of creating the dataset
For reproduction of DPO (direct preference optimization) thesis experiments(https://arxiv.org/abs/2305.18290)
2. How data is produced
To reproduce the paper's experimental results, we need (x, chosen, rejected) data.However, imdb data only contains good or bad reviews, so the data must be readjusted.
2.1 prepare imdb… See the full description on the dataset page: https://huggingface.co/datasets/insub/imdb_prefix20_forDPO_gpt2-large-imdb-FT_siebert_sentiment-roberta-large-english.imdb_prefix3_forDPO_gpt2-large-imdb-FT_siebert_sentiment-roberta-large-english
Dataset Card for "imdb_prefix3_forDPO_gpt2-large-imdb-FT_siebert_sentiment-roberta-large-english"
More Information needed
twitter-100m-siebert-activations