# Autoregressive Generation

Janus-Pro is a novel autoregressive framework that unifies multimodal understanding and generation. It processes multimodal tasks using a single unified Transformer architecture by decoupling visual encoding paths.

Yi-Ko-6B is an advanced version of the 01-ai/Yi model, further pre-trained with an extended vocabulary and Korean/English corpus, supporting bilingual text generation in Korean and English.

Goliath 120B is an autoregressive causal language model created by merging two fine-tuned Llama-2 70B models, supporting conversational tasks.

A large-scale molecular generation model based on the SELFIES molecular language, capable of de novo molecular generation or completing partial molecular structures.

MusicGen is a text-to-music model that generates high-quality music samples based on text descriptions or audio prompts, utilizing a 1.5-billion-parameter autoregressive Transformer architecture.

LLaMA is an efficient foundational language model developed by Meta AI, available in parameter sizes ranging from 7B to 65B. Based on the Transformer architecture, it is suitable for various natural language processing tasks.

Donut is an OCR-free document understanding Transformer model that combines a visual encoder and text decoder for image-to-text conversion

Donut is an OCR-free document understanding Transformer model composed of a visual encoder (Swin Transformer) and a text decoder (BART).

s2t-small-librispeech-asr is a speech-to-text (S2T) model for automatic speech recognition (ASR), based on a sequence-to-sequence transformer architecture.

Boris is an autoregressive language model based on the GPT-J architecture with 6 billion parameters, specializing in French text processing.