Closed drleehw closed 1 year ago
Hi. I can't fully replicate your example as I don't have the our_date_data
object.
I'd do it like this: create a new variable that has the prompt in it and paste it to the tweet of interest -->
tweet_data$prompt = paste0('Categorise this tweet whether the tweeter would be happy or not.
Respond with just [positive/negative/neutral]. \nTweet: ', tweet_data$tweets, '\n Response: ')
GPT_tweet_sentiment <- gpt3_completions(prompt_var = tweet_data$prompt,
id_var = tweet_data$promptID,
param_model = 'text-davinci-003',
param_max_tokens = 100,
param_n = 1,
param_temperature = 0.4)
Output (no rate limit issues) is in the expected form (i.e., negative, positive, ...).
Could you try running this? If that doesn't work, check if you're still within your API limits (funds and/or expiry of free trial). There are some limits at play but I don't think these apply for this example.
Let me know if this worked or not.
I've tried playing with various parameters, but I seem to be limited to just 10 responses. Is this a limitation or am I doing something stupid?