FunGPT - an open-source AI sentiment regulation project based on the InternLM2.5 series models.
FunGPT is an open-source project based on the InternLM2.5 series of large models, specifically designed for emotion regulation. It features two core functions: a sweet talk mode and a sharp criticism mode. The sweet talk mode uses warm words and unique...
What is FunGPT?
FunGPT is an open-source project based on the InternLM2.5 series of large models, specifically designed for emotion regulation. It features two core functions: a "sweet talk" mode and a "sharp criticism" mode. The sweet talk mode uses warm words and unique praise to improve the user's mood, while the sharp criticism mode helps users release stress in a humorous and engaging way. FunGPT uses the 1.8B series of lightweight models, combined with AWQ quantization technology, which saves GPU memory and improves inference speed.
FunGPT's main functions
- Sweet talk modeWhen users are feeling down, the mode can instantly boost their mood. Master Zan will praise users in the most appropriate and unique way, making users' confidence soar.
- Sharp and sarcastic remarks modeWhen users feel overwhelmed by stress, this mode allows them to find an outlet by roasting others. The roaster's words are not only sharp but also humorous, allowing users to experience creative and imaginative ways to roast others.
- Lightweight modelThe 1.8B series of lightweight models has been released, featuring a smaller size and excellent performance. It employs AWQ quantization technology, saving GPU memory while improving inference speed.
FunGPT's technical principles
- Model ArchitectureFunGPT uses the InternLM2.5 series of models as its basic architecture. Based on the Transformer architecture, the model possesses powerful language generation and understanding capabilities. The core advantage of the Transformer architecture lies in its multi-head attention mechanism, which can examine text from different perspectives and capture long-distance dependencies and contextual information.
- Fine-tuning technologyTo meet personalized user needs, FunGPT uses Xtuner for instruction and full fine-tuning. This allows the model to better adapt to specific task scenarios, such as a sweet talk mode and a sharp retort mode.
- Quantitative techniquesFunGPT employs AWQ (Adaptive Weight Quantization) technology. By reducing the storage space of model parameters, it lowers the barrier to entry for using the model, improves inference speed, and enables the model to run efficiently on resource-constrained devices.
FunGPT project address
- Github repository:https://github.com/Alannikos/FunGPT
FunGPT Application Scenarios
- Creative inspirationWhen you need new creative inspiration, FunGPT can help you generate interesting ideas, such as providing inspiration for writing, drawing, or design.
- Entertainment and leisureWhen users are bored, FunGPT can recommend entertainment content such as movies, music, and books, or engage in interesting conversations with users to enrich their leisure time.