Reputation: 378
Usually, one can upload config data to wandb init like this:
MODEL_CONFIG = {
"num_epochs": config.num_epochs,
"batch_size": config.batch_size,
"base_lr": config.base_lr
}
wandb.init(config=MODEL_CONFIG)
The problem is that, when I do a sweep, like in the collab example:
sweep_config = {
"method": "bayes",
"metric": {"name": "accuracy", "goal": "maximize"},
"parameters": {"some_other_parameter": {"values": [1, 2, 3]}}
sweep_id = wandb.sweep(sweep_config, project="please-help")
def train(config=None):
# Initialize a new wandb run
with wandb.init(config=config):
# If called by wandb.agent, as below,
# **this config will be set by Sweep Controller**
config = wandb.config
...
wandb.agent(sweep_id, train, count=5)
This config parameter gets phagocyted by the sweep mechanism, and I can't use it to upload extra information. I know that I could add extra info in a table or something like this, but I really like how tidy it is when in the "Overview" page.
Am I missing something? Is there a way around this?
Upvotes: 0
Views: 23