For example, analzying plant images like this one – www.kaggle.com/c/plant-pathology-2021-fgvc8/data
How much computing power would it cost to train my model, given that I would know what I’m doing, to place in the top 10% of tournament results?
I know training things like GPT3 costs millions, but I’m just wondering what the costs for computing power are needed to train models in smaller competitions such as these. I’m not referring to the competition where you use kaggle’s own computing resources and are free to compete in. I’m more wondering about the financial cost of competing in tournaments that have resource intensive datasets that range into the 10-20GB of data range and can’t be done on a low specced computer.
I’m really interested in this information as it relates to kaggle competitions specifically.
Or if you work in the Machine Learning field, some references to the financial costs of training the algorithms would be nice, especially if the dataset amounts are similar to those of Kaggle tournaments.
Not interested in the manpower and the price of the data itself, just the cost of computing power for training the models. Thanks.
Read more here: Source link