Hello pytorch community. I am a beginner, and I have used nn.Sequential and nn.RNN to fit timeseries data I generated in a special form and I obtain different performance (both for mean square error and for mean absolute relative error). I would like to gather a maximum amount of plot ideas to realize either using pytorch packages (any recommendations?) or matplotlib / plotly to show the details of the performance of each model (I might add more in the future) in order to not only get a good idea of how well it performs but also the comparison in performance. Any other metric ideas would be greatly appreciated as well. Thank you.
Hello @Kenza_Benjelloun
When it comes to generating plots to compare metrics of different models, I would recommend the Weights&Biases package. You can follow this guide to get started with it.
Essentially, you’re can initialize it in your Python script like this:
run = wandb.init(
project="timeseries",
config={
"model": "sequential"
},
)
And then you can log data to wandb, which can be your metrics:
wandb.log({"mse": mse, "mae": mae})
This will automatically generate plots that are viewable in your wandb projects online, allowing you to compare results across different models and experiments, combine plots, and watch data being generated in real time.
Machine Learning Engineer at RidgeRun.ai
Contact us: support@ridgerun.ai