Creating Datasets is Hard
Sorting all of your examples into individual entries takes a ton of time. Formatting the text takes a ton of time.
Riku makes it fun. Be excited about creating your next dataset.
Our JSONL dataset builder is super simple to use and will perfectly format your dataset ready for fine-tuning. Enter both your prompt and completion with ease and download your JSONL file at any time.
Datasets can come in all shapes and sizes. We advise you to provide atleast 50 examples. For more complicated tasks and processes, the general train of thought is; 'more is better', up to thousands of examples.
Fine-tuning in Riku
Using a dataset for a fine-tuned models escapes the prompt token limits of a raw large language model. 2048 tokens is approximately 8,000 characters. Using datasets removes that hard limit and gives you better content. Win, win!
