Instrumenting W&B in your code

Key Takeaways

to get started the first thing we need to do is install the python client wand B we can do so by calling pip install wand B after you've created an account you can log in locally by calling wandbead.login in python or calling 1B login from a terminal you can then follow the link to get your API key and paste it in the Box shown this is the same code as in our previous video where we're using print to log our metrics and our config values are hard coded this is that same training script but slightly refactored to use weights devices to tell weights and biases that we'd like to begin to capture details of the code we're running we want to tell it to create a run a run is how weights and biases organizes the details of every time that you run your code to create a run we just call wand B dot init we've also refactored this code to gather all of our configs at the top of the script and then pass that to 1b.init we're also referencing that same config whenever we're using those values throughout our script here is that same script but refactored even further so that rather than printing our epochs and our loss thread training we're passing it to wand b.log in a dictionary this is where we put metrics from our run like the model training or validation loss and we can call this many times when running the script and weights and biases will store and visualize the history of each of the different metrics as we log them once we've done all this we're a long way towards having reproducible experiments in the next video I'll run this script and I'll show you how we can get insights from the weights and biases workspace