How to live log arbitrary line graph

I’m running active learning experiments where the result is some model performance metric against dataset size. I’ve seen the below code example:

data = [[x, y] for (x, y) in zip(x_values, y_values)]
table = wandb.Table(data=data, columns = ["x", "y"])
wandb.log({"my_custom_plot_id" : wandb.plot.line(table,
                                 "x", "y", title="Custom Y vs X Line Plot")})

but this seems to be for static graph generation once you have access to all the data.

How do I generate a live updating graph of model performance against dataset size?

Hi @georgepearsebehold,

Just to make sure I understand you correctly, are you trying to log a line of values at each timestep and want to see the progression of this line over time?


I wanted to log a line of values against a self defined x axis.

I’ve since resolved with :

wandb.define_metric("val/ AUC SLC (max)", step_metric="dataset_size")
log_dict = {
    'dataset_size': len(baal_data_module.active_set),
    'val/ AUC SLC (max)': max_val_auc

Ah, understood. define_metric is definitely the right way to go about this. Glad you were able to resolve this by yourself!