Improving turkish grammar and reasoning capabilities of open source large language models with parameter efficient fine tuning

Large Language Models have recently gained a lot of attention following the GPT-3 model. Then different open source language models were developed. With the development of these, the fine-tuning of large language models has become of great importance. One of the most important issues that has been the subject of recent research is fine-tuning in an efficient and cost-effective way. There are some problems when fine-tuning Large Language Models. Examples of these problems are the problem of insufficient storage space due to the large parameter size, the speed problem due to the large size, etc. Different approaches have been and continue to be developed to find solutions to these problems. In this paper, we present techniques for fine-tuning large language models and techniques that can be applied for speed, efficiency and cost reduction in storage. These techniques are different Parameter-Efficient Fine Tuning techniques. The main ones are prompt tuning, Low-Rank Adaption(LORA), adapters and prefix tuning.