1 of 8

SmartSpace LLMs

|Vedant Talati | Sathvik Samant |

| Jose Rubio | Jonathan Duran |

| Advisor: Dr. Ortiz |

2 of 8

Research Paper

  • Presenting a research paper to Dr. Ortiz and team → WARM: On the Benefits of Weight Averaged Reward Models

​

  • Novel method to develop + train reward models → Significant efficiency gains

​

  • Can be implemented in this project for the LLM → provide reliable, human-aligned responses

​

  • Annotations + Slideshow

3 of 8

Loguru Logging

  • Used Loguru library for logging script output
  • Organized, timestamped way to log and debug
    • logger.error(‘error message’)
  • Added function to download and clear old logs when they aren’t necessary anymore
    • .truncate(0)

​

4 of 8

Sensor Testing

  1. Tested the following sensors:
    1. Accelerometer
    2. Humidity
    3. Temperature
    4. Air Quality
    5. Infrared Motion Sensing
    6. RGB Values
  2. Still not fully working:
    • Audio
      1. Hertz or Decibels?
      2. Frequency? Volume? Voltage?

5 of 8

Frontend development

  • New chart feature added to display sensor data that is coming through while running experiment
  • Chart feature integrated into shutdown Pi feature and both can now work simultaneously

​

​

6 of 8

Github

  • Finished all installation of dependencies on a Maestro and documentation
    • understand how the working Maestros came to be
    • have a reference on how to create your own working Maestro for experimentation.

7 of 8

  1. Create a Visual understanding for Active Maestros that are streaming data and in experimentation
  2. Be able to have request methods on the API for the different feature sin the frontend
  3. Begin ML

PLANS FOR NEXT WEEK

8 of 8

Thank you