MindedWheeler

Embody_AI with car as Demo

MindedWheeler

🌐 Website β€’ πŸ€— Model

🌈 Update

  • [2024.02.23] πŸŽ‰πŸŽ‰πŸŽ‰ MindedWheeler is publishedοΌπŸŽ‰πŸŽ‰πŸŽ‰

πŸ€– Model Training Data

UserοΌšεΏ«ι€Ÿε‘ε·¦θ½¬
RobotAI: (1.0, -0.3)    
...
  • The two float are in range [-1,1]
  • The first float is speed, the second is direction (negative means left, positive means right).

πŸ€– Communication Protocol

  • 0x02, 0x02, 0x01, 8, data_buf; (See detail in code)

ℹ️ Usage

  1. DownLoad πŸ€— Model get model.bin.

    cd MindedWheeler
    git submodule update --init --recursive
    python qwen_cpp/convert.py -i {Model_Path} -t {type} -o robot1_8b-ggml.bin
    

    You are free to try any of the below quantization types by specifying -t :

    • q4_0: 4-bit integer quantization with fp16 scales.
    • q4_1: 4-bit integer quantization with fp16 scales and minimum values.
    • q5_0: 5-bit integer quantization with fp16 scales.
    • q5_1: 5-bit integer quantization with fp16 scales and minimum values.
    • q8_0: 8-bit integer quantization with fp16 scales.
    • f16: half precision floating point weights without quantization.
    • f32: single precision floating point weights without quantization.
  2. Install package serial.tar.gz

     cd serial
     cmake .. & make & sudo make install
    
  3. Compile the project using CMake:

    cmake -B build
    cmake --build build -j --config Release
    
  4. Now you may chat and control your AI car with the quantized RobotAI model by running:

    • qwen.tiktoken is in the model directory
    ./build/bin/main -m robot1_8b-ggml.bin --tiktoken qwen.tiktoken -p θ―·εΏ«ι€Ÿε‘ε‰
    

    To run the model in interactive mode, add the -i flag. For example:

    ./build/bin/main -m robot1_8b-ggml.bin --tiktoken qwen.tiktoken -i
    

    In interactive mode, your chat history will serve as the context for the next-round conversation.

πŸ₯Έ To do list

  • Continue to create data and train a robust model
  • Add ASR and TTS
  • ...

✨ Citation

Please use the following citation if you intend to use our dataset for training or evaluation:

@misc{MindedWheeler,
  title={MindedWheeler: Embody_AI with car as Demo},
  author={Xidong Wang*, Yuan Shen*},
  year = {2024},
  publisher = {GitHub},
  journal = {GitHub repository},
  howpublished = {\url{https://github.com/FreedomIntelligence/MindedWheeler}},
}

πŸ€– Acknowledgement

Downloads last month
11
Safetensors
Model size
1.84B params
Tensor type
BF16
Β·
Inference Examples
Inference API (serverless) does not yet support model repos that contain custom code.