| ben-aaron188 | 3818e7c | 2022-09-08 17:49:01 +0200 | [diff] [blame^] | 1 | gpt3.bunch_request = function(data |
| 2 | , prompt_var |
| 3 | , completion_var_name = 'gpt3_completion' |
| 4 | , param_model = 'text-davinci-002' |
| 5 | , param_suffix = NULL |
| 6 | , param_max_tokens = 256 |
| 7 | , param_temperature = 0.9 |
| 8 | , param_top_p = 1 |
| 9 | , param_n = 1 |
| 10 | , param_stream = F |
| 11 | , param_logprobs = NULL |
| 12 | , param_echo = F |
| 13 | , param_stop = NULL |
| 14 | , param_presence_penalty = 0 |
| 15 | , param_frequency_penalty = 0 |
| 16 | , param_best_of = 1 |
| 17 | , param_logit_bias = NULL){ |
| 18 | |
| 19 | |
| 20 | data_ = data |
| 21 | |
| 22 | data_length = data_[, .N] |
| 23 | |
| 24 | data_[, completion_name := ''] |
| 25 | |
| 26 | |
| 27 | for(i in 1:data_length){ |
| 28 | |
| 29 | print(paste0('Request: ', i, '/', data_length)) |
| 30 | |
| 31 | row_outcome = gpt3.make_request(prompt = as.character(unname(data_[i, ..prompt_var])) |
| 32 | , model = param_model |
| 33 | , output_type = 'detail' |
| 34 | , suffix = param_suffix |
| 35 | , max_tokens = param_max_tokens |
| 36 | , temperature = param_temperature |
| 37 | , top_p = param_top_p |
| 38 | , n = param_n |
| 39 | , stream = param_stream |
| 40 | , logprobs = param_logprobs |
| 41 | , echo = param_echo |
| 42 | , stop = param_stop |
| 43 | , presence_penalty = param_presence_penalty |
| 44 | , frequency_penalty = param_frequency_penalty |
| 45 | , best_of = param_best_of |
| 46 | , logit_bias = param_logit_bias) |
| 47 | |
| 48 | |
| 49 | data_$completion_name[i] = row_outcome$choices[[1]]$text |
| 50 | |
| 51 | |
| 52 | } |
| 53 | |
| 54 | data_cols = ncol(data_) |
| 55 | names(data_)[data_cols] = completion_var_name |
| 56 | |
| 57 | return(data_) |
| 58 | } |