Currently, when I run multiple experiments/large batch sizes, sometimes it encounters rate limits soon, sometimes not. I wonder if the rate limit strategy could be more transparent so that I can adjust my hyperparameters accordingly. Or whether the API can return a rate limit exception so that I can sleep() before submitting the request again.
Currently, when I run multiple experiments/large batch sizes, sometimes it encounters rate limits soon, sometimes not. I wonder if the rate limit strategy could be more transparent so that I can adjust my hyperparameters accordingly. Or whether the API can return a rate limit exception so that I can sleep() before submitting the request again.