T5 V1 1 Base Ko

🚀 team-lucid/t5-v1_1-base-ko

This is a Korean-trained version of Google's T5 (Text-to-Text Transfer Transformer) Version 1.1, leveraging a Korean corpus for training.

This model, t5-v1_1-base-ko, is a T5 v1.1 model trained on a Korean corpus. To prevent Out-of-Vocabulary (OOV) issues, Byte-level Byte Pair Encoding (BBPE) was employed. After observing that morphological analysis improved performance in HyperCLOVA, MeCab was used during the tokenizer training process to ensure that morphemes are not tokenized incorrectly.

This research was trained on Cloud TPUs supported by Google's TPU Research Cloud (TRC).

🚀 Quick Start

This section provides a brief introduction to getting started with the team-lucid/t5-v1_1-base-ko model.

💻 Usage Examples

Basic Usage

from transformers import AutoTokenizer, T5ForConditionalGeneration

tokenizer = AutoTokenizer.from_pretrained('team-lucid/t5-v1_1-base-ko')
model = T5ForConditionalGeneration.from_pretrained('team-lucid/t5-v1_1-base-ko')

📄 License

This project is licensed under the Apache-2.0 license.