bilibot

A local chatbot trained by fine-tuning Bilibili user comments

CommonProductchattingChatbotLocal Deployment

bilibot is a local chatbot trained on Bilibili user comments, supporting both text chat and voice dialogue. It uses Qwen1.5-32B-Chat as the base model and is further fine-tuned with Apple's mlx-lm LORA project. The voice generation part is based on the GPT-SoVITS project, utilizing the Paimon voice model. This chatbot can quickly generate conversational content and is suitable for scenarios requiring intelligent dialogue systems.

Supports text chat to meet basic communication needs.
Can generate voice dialogues in response to given questions
providing richer interaction experiences.
Utilizes Qwen1.5-32B-Chat as the base model
ensuring high-quality and diverse dialogues.
Fine-tuned through training
making the chatbot closer to Bilibili users' communication habits.
Employs GPT-SoVITS for voice generation
enhancing the naturalness of voice interaction.
Model quantization compression accelerates content generation speed

bilibot is suitable for developers and enterprises that need intelligent chat systems
especially those who want to provide personalized services and enhance user experience. This includes fields like customer service
online education
and entertainment interaction.

As an online customer service representative
providing 24/7 consultation services.
As a smart tutor on an educational platform
assisting with teaching and answering questions.
Providing role-playing and interactive experiences in entertainment applications.

First
visit the bilibot GitHub page and clone or download the project code.
Configure the Python environment according to the project documentation. Anaconda is recommended.
Run the model fine-tuning training and inference testing. Merge the fine-tuned adapters file with the base model.
Use quantization acceleration tools to compress the model to improve generation speed.

Visit

bilibot Visit Over Time

Monthly Visits

499904316

Bounce Rate

37.31%

Page per Visit

5.8

Visit Duration

00:06:52

bilibot Visit Trend

bilibot Visit Geography

bilibot Traffic Sources

bilibot Alternatives

IncarnaMind — Engage in conversations with your documents using large language models like GPT.

Productivity•Machine Learning•PDF

120

CosyVoice — A multilingual large-scale voice generation model, providing full-stack capabilities for inference, training, and deployment.

Others•Multilingual•Voice Generation

2964

FlyCode — Utilize AI to optimize subscription revenue and reduce customer churn.

Business•Subscription Revenue Optimization•Payment Retry

102

DeepLearning.AI — A platform offering professional courses and resources in the field of AI.

InternationalSelection•Machine Learning•Deep Learning

GVHMR — Human motion recovery based on gravity perspective coordinates to restore world positioning.

Image•Human Motion Recovery•Computer Vision

JanitorAI — A platform offering a diverse range of chatbots.

Copilot in Excel — Copilot in Excel: Unlocking data-driven decision-making.

Productivity•Data Analysis•Automation

OmniGen — A unified framework for image generation that simplifies multi-task image generation.

Image•Image Generation•Diffusion Models

Mistral-Small-Instruct-2409 — High-performance instruct-tuned AI model

Programming•Artificial Intelligence•Natural Language Processing

Qwen2.5-LLM — An open-source high-performance language model that supports multi-platform applications.

Productivity•Natural Language Processing•Machine Learning

ViewCrafter — Video diffusion model for high-fidelity new viewpoint synthesis

Image•3D Reconstruction•Video Synthesis

xAI — AI to Accelerate Human Scientific Discovery

DeepMind — A leading artificial intelligence research company under Google

1X World Model — An advanced world model providing virtual simulation and evaluation for robotics.

Productivity•Machine Learning•Virtual Simulation

Pixtral 12B — The first multimodal Mistral model, supporting hybrid task processing for images and text.

Productivity•Multimodal•AI Model

Shangchen Zhou — A blog website focused on research and innovation in the fields of computer vision and machine learning.

OpenSource•Computer Vision•Machine Learning

Skywork-Reward-Llama-3.1-8B — An advanced reward model for text classification and preference assessment.

Programming•Machine Learning•Natural Language Processing

seed-vc — Zero-shot voice conversion technology that achieves high-fidelity transformation of quality and tone.

Programming•Voice Conversion•Zero-shot Learning

114

pixtral-12b-240910 — A multimodal large language model that supports understanding of both images and text.

Image•Multimodal•Image Processing

MemoRAG — A memory-based RAG framework for general-purpose applications

Programming•Natural Language Processing•Machine Learning

finegrain-object-cutter — Fine-grained object cutting tool for precise image editing.

Image•Image Editing•Machine Learning

Qihoo-T2X — Qihoo-T2X, an efficient diffusion transformer model aimed at text-to-any-task processing.

Programming•Text Processing•Diffusion Transformer

SuperPrompt — Aimed at helping us understand the engineering of AI agents through prompt design.

Programming•Artificial Intelligence•Machine Learning

rStar — Enhances problem-solving capabilities of small language models through self-play mutual reasoning.