Hi
I am very new to GPT-3 and have been using it for two days only.
The reason why I am using GPT-3 is to create a classifier.
The API CURL calls to a classifier include the “examples”.
However, the classical approach to creating a classifier is to provide the training data, create a model, and then simply execute the query on the model.
The pricing is very important, and in order to avoid getting over-charged for the examples I will be using over-and-over again, I would like to somehow train the GPT-3 classifier and to simply make the next REST call as a query (without a necessity to provide the examples).
Is such a thing possible to be done, or do I really need to provide the examples each time I make the REST call to API ?
Thank you kindly for your answer, and I apologise if this question has been asked/answered somewhere else… I am still very new.
EDIT: I see the option to upload the file, but I do not see how we get charged when searching the file or per file content/size.
@damir.olejar Please take a look at our Pricing Page for more info on how pricing is calculated for endpoints that use files i.e. Answers, Search, and Classifications.
1 Like
Already did:
…In this scenario, costs are largely based on the number of examples reranked (controlled by max_examples ) and the total length of those examples. If you pass examples in your request instead, costs are based on the total length of all those examples…
Does not state HOW the prices are calculated per file length, the number of examples, etc. The problem is, I need exact numbers and not a description.
Thank you for replying
On that page, please see the FAQ about Search. It has a formula that you can use to estimate tokens. The answers and classification use a combination of search + completion, so it should give you a ballpark. You can also try out a few examples with the ‘ada’ engine (our fastest and least expensive engine → $0.0008/1k tokens) to test how many tokens are used.
1 Like
Thank you. I will also track the usage and do the estimate.
1 Like