banksy23

Code-Eval

Evaluate the code generation ability of Code LLMs. Supports mainstream benchmarks such as HumanEval and MBPP. Define chat templates and pre/post-processing code in interface form to facilitate customizing the logic for extracting valid code snippets for testing for one's own model.

GALAXY

Official repository of the AAAI'2022 paper "GALAXY: A Generative Pre-trained Model for Task-Oriented Dialog with Semi-Supervised Learning and Explicit Policy Injection"

MTTOD

NavigationInCollege

P-tuning

A novel method to tune language models. Codes and datasets for paper ``GPT understands, too''.

P-tuning-v2

An optimized deep prompt tuning strategy comparable to fine-tuning across scales and tasks

PrefixTuning

Prefix-Tuning: Optimizing Continuous Prompts for Generation

ToD-BERT

Pre-Trained Models for ToD-BERT