I have about 1400 lines like these below but preliminary tests on the trained model are not satisfactory. Is it due to some weakness in the structure of these prompt-completion pairs or to the possibly bad prompt I may use to test the model?
{"prompt": "Use formula function signature COUNT(list: Array) that returns the number of items in the given list. The data source is: [1, 2, 3], and the expected return result is: 3.\n\n###\n\nSuggestion:", "completion": " COUNT([1, 2, 3])###"}
{"prompt": "Use formula function signature COUNT(list: Array) that returns the number of items in the given list. The data source is: [], and the expected return result is: 0.\n\n###\n\nSuggestion:", "completion": " COUNT([])###"}
{"prompt": "Use formula function signature DIFFERENCE(sourceList: Array, items_to_exclude: Array) that returns a list with all the values from the first list that are not present in the second list. In other words, excludes all values from the first list that are given as a second list. The data source is: numbers1 = [10,55,66,54,85,91] and numbers2 = [52,55,85,10,94], and the expected return result is: [66,54,91].\n\n###\n\nSuggestion:", "completion": " DIFFERENCE(numbers1, numbers2)###"}
{"prompt": "Use formula function signature DIFFERENCE(sourceList: Array, items_to_exclude: Array) that returns a list with all the values from the first list that are not present in the second list. In other words, excludes all values from the first list that are given as a second list. The data source is: numbers1 = [10,55,66,54,85,91] and numbers2 = [52,55,85,10,94], and the expected return result is: [52,94].\n\n###\n\nSuggestion:", "completion": " DIFFERENCE(numbers2, numbers1)###"}