Hi, I’m recently working with a multi-agent learning environment with pytorch. I tried to log using Tensorboard with tensorboard --logdir=xxx
, but this doesn’t show any logged metrics. I then tried to use a wandb logger by doing this:
from ray.tune.integration.wandb import WandbLogger
# Some setup code
run_config["logger_config"] = {
"wandb": {
"project": "name1",
"api_key_file": "./key.txt"
}
}
ray.init()
results = tune.run("PPO", config=run_config, loggers=[WandbLogger])
ray.shutdown()
In the wandb logger website only system metrics such as GPU util are logged, and the hyperparameters are all properly logged. What might be the problem here?