W&B Models by Weights & Biases
Train, fine-tune, and manage AI models
Start new thread
Stephanie Sher

Weights & Biases - Developer tools for deep learning & machine learning

by•
Easily track & save the performance of your deep learning models. In < 5 lines of code, integrate Keras, Tensorflow, Pytorch, or Fastai models so that every hyperparameter & metric is saved each time code runs.

Add a comment

Replies

Best
Lachlan Kirkwood
Looks like such a powerful tool for extracting performance insights! Absolutely love the documentation feature, awesome work!
Carey Phelps
@lachlankirkwood Thanks! So delighted to hear you like it :)
Bona
This is awesome and so Any discounts or special pricing for researchers/students/non-professionals?
Carey Phelps
@gamblers_united Yes, we offer free accounts for academics and open source projects.
Steve Goolsby
Amazing. I think this is very helpful tools for us. Keep it up & go ahead.
Carey Phelps
@stevegoolsby Thanks so much! Sign up at app.wandb.ai for a free account to check it out :)
Carey Phelps

This simple system of record automatically saves logs from every experiment, making it easy to look over the history of your progress and compare new models with existing baselines.

Pros:

Easy, fast, and lightweight experiment tracking

Cons:

Only available for Python projects

Alex Conway
Very cool! I hacked together something similar but much more basic for personal use and always wondered why TensorBoard didn't solve this problem. I just wish this was open source! :) P.S. awesome use of the parallel co-ordinates d3.js chart - great idea to apply it to experiment configurations!