[samzer]

Hi,

The author here.

ModelChimp is an experiment tracker for Deep Learning and Machine Learning experiments.

ModelChimp provides the following features:

  - Real-time tracking of parameters and metrics
  - Realtime charts for experiment metrics at epoch level
  - Code used for the experiment
  - Experiment comparison
  - Collaborate and share experiments with team members
  - Python objects storage such as data objects and model 
    objects which can be used pulled for other experiments
  - Storage of test and validation images for computer vision 
    use cases. Useful for post experiment forensics of deep 
    learning models
  - Server based solution with user registration and 
    authentication

[yodon]

This looks very cool. Do you see it as tightly coupled to ML or would you think this is adaptable to more general tracking of dataset processing? It looks pretty flexible to me after a quick glance but I haven't had time to dig deeper and understand the details of how it works.

[samzer]

I started off having strong integration with ML frameworks but eventually, I made it more generic to track any kind of metrics and parameters especially if one is using a custom algorithm.

I'm working on adding data versioning where one can keep version their datasets and pull it anywhere using a line of code in python or command line.

[yodon]

Very cool. Is there a way to add thumbnails or similar to the datasets/UI, to help someone who is trying to find the right dataset spot it more easily without having to go in and open individual files?

[samzer]

Yeah, I'll be adding that. There will be an option to toggle between list and thumbnail view.