With a unified and flexible framework, it supports score-to-song conversion, controllable generation, music and lyrics editing, and low-threshold voice cloning, etc.

Decoupling control flow and computing flow for both flexibility and efficiency.

Comprehensive efforts have been made in the research and development of the basic models, which have unique advantages in various business scenarios.

ByteCheckpoint improves performance by up to 529.22 times in saving and 3.51 times in loading speeds.

Doubao large model's daily token usage exceeds 50 bn.

Volcano Engine unleashes the power of large models to support academic frontier exploration

Generative AI further enables the integration between the industry and the academic and research fields.

The first author of the paper is a team intern.

The relevant technology has been applied for a while and has been recognized by users in real-life scenarios.

This paper proposes a novel IR-QLoRA for pushing quantized LLMs with LoRA to be highly accurate through information retention. This is the first time the perspective of information theory has been introduced and theories related to information entropy are utilized to examine and measure the quantization of large models. The paper has been selected for an oral presentation at ICML 2024. The first author of the paper is an intern with the speech group of ByteDance's Doubao Team (Seed) and a candidate in the ByteDance Scholars Program. Teachers and students from the State Key Laboratory of Complex & Critical Software Environment at Beihang University also contributed to the research.