In 2024 World Artificial Intelligence Conference(WAIC 2024), the Beeper ("Station B") announced a number of self-developed AI technology achievements and AIGC diversified ideas, including the latest customized AI voice sound library, self-developed audio and video big model must cut Studio and self-developed AI dynamic diffusion technology.
also,B Station's own researchLarge Language ModelThe series is also on display for the first time at WAIC 2024.It includes the open source Index-1.9B chat and Index-1.9B character models.
Query GitHub to learn, Index-1.9B series of models in June open source, including the base model, control group, dialog model, role-playing model:
-
Index-1.9B base : Base model with 1.9 billion non-word embedded parameters, pre-trained on 2.8T Chinese and English dominated corpus, leading the class on multiple benchmarks.
-
Index-1.9B pure : A control group for the base model with the same parameters and training strategy as base, with the difference that all instruction-related data in this version of the corpus was filtered to verify the effect of instructions on the benchmark
-
Index-1.9B chat : Dialog model based on index-1.9B base aligned by SFT and DPO
-
Index-1.9B character : Introduced RAG for fewshots role-playing customization based on SFT and DPO
At the 15th anniversary speech of B Station, Chen Rui, Chairman and CEO of B Station, said that in 2023, the average daily video playback of AI-related content on B Station will grow by more than 80% year-on-year, and the explosive content will cover the fields of popular science information, AI technology applications, digital people and creative applications.
According to the data revealed by the B station, more than 80 million users watch AI-related videos on the B station every month, of which 60% are post-00s.
Index-1.9B series model open source address:
https://github.com/bilibili/Index-1.9B