Failing to fine-tune a model, invalid jsonl format no matter what

[ruby_coder]

@Arivald

I wonder if it perhaps dislikes my environment. I’m on an older version of Win 10, creating the file via VSCode and running the fine-tuning via Git Bash.

My experience is that it is best (less errors and headaches) to use the full path to the file in your File API call.

/the/full/path/to/your/file/data_train.jsonl

The API error message are beta, hit or miss (mostly miss), so I discovered this “the hard way”.

Here is how I do it…as an example:

module Files
    def self.get_client
        Ruby::OpenAI.configure do |config|
            config.access_token = ENV.fetch('OPENAI_API_KEY')
        end
        client = OpenAI::Client.new
    end


    def self.upload(filename="#{Rails.root}/app/assets/files/fine-tune.jsonl",purpose='fine-tune')
        client = get_client
        response =client.files.upload(
            parameters: {
                    file: filename,
                    purpose: purpose,
                })
        file_id = JSON.parse(response.body)["id"]
        file_id 
    end
end

Footnote

Please note @Arivald, even if your data passes JSONL validation, you must use a specific format in your text for fine-tuning. Your current JSONL text will pass JSONL but it will does not “pass” (there is no validator for this, but I have one), for the API requirements.

Most Developers Seem to Miss this Requirement

Reference:

Preparing You Dataset

Here is an example of a fully “API Validated” entry. Note I use “PROMPT_SEPARATOR” and “STOPSTOP” in this example. You can choose whatever you like.