Jaewoo (Kyle) Song

gpt2-dialogue-generation-pytorch. The PyTorch implementation of fine-tuning the GPT-2(Generative Pre-trained Transformer 2) for dialogue generation.

transformer-translator-pytorch. The PyTorch implementation of the transformer for machine translation.

bert-crf-entity-extraction-pytorch. Entity extraction using BERT + CRF for single-tun / multi-turn setting in dialogues

lstm-bayesian-optimization-pytorch. Bayesian Optimization implementation for text classifiction

recosa-dialogue-generation-pytorch. The PyTorch implementation of ReCoSa(the Relevant Contexts with Self-attention) for dialogue generation using the multi-head attention and GRU.

t5-dst-modified-pytorch. Modified version of T5-DST for Dialogue State Tracking.

dialogue-sentence-bert-pytorch. DialogueSentenceBERT: SentenceBERT for More Representative Utterance Embedding via Pre-training on Dialogue Corpus

labyrinth-simulator-kani. This is an open-sourced project for simulating the fantasy text-based adventure game: Jim Henson's Labyrinth: The Adventure Game grounded on large language models supported by kani.

dqn-atari-tensorflow. This is a refactored version of Deep Q-Network for Atari.

dialogue-error-correction-pytorch. Dialogue Error Correction for Dialogue State Tracking and Data Augmentation.

algorithm-study. Practice for Algorithm with C++

gpt2-instruction-tuning-rlhf-pytorch. PPO (Proximal Policy Optimization) and DPO (Direct Preference Optimization) practice code to train a GPT-based model that generates a response for the given task-specific instruction.