README.md

metadata

license: cc

Bert2Bert (Encoder-Decoder) on Liputan6 100k dataset

Dataset source: https://huggingface.co/datasets/fajrikoto/id_liputan6
Model used for Fine Tuning (Encoder-Decoder):
https://huggingface.co/cahya/bert-base-indonesian-1.5G

Trained on 1x3090 @ 8 epoch (EarlyStopping Callbacks)

Train logs, metrics, and params: https://wandb.ai/willy030125/huggingface/runs/sb2kcuck
https://www.comet.com/willy030125/huggingface/5dd7c19d0c85472abdf4136529f4322c
Eval results and Perplexity: eval_results.json

Usage:

from transformers import AutoTokenizer, EncoderDecoderModel
tokenizer = AutoTokenizer.from_pretrained("Willy030125/Bert2Bert_Liputan6_100k_8epoch")
model = EncoderDecoderModel.from_pretrained("Willy030125/Bert2Bert_Liputan6_100k_8epoch")