Seeking Advice: Fine-Tuning GPT Models - OpenAI Developer Forum
As I near the completion of my research dissertation as a Master’s student, I am delving into the realm of fine-tuning pre-trained GPT models for off-label applications. Specifically, my project focuses on utilizing these models in the financial stock market space, where I am collecting daily news and stock prices for analysis, primarily centered around the SPY index. To aid in this endeavor, I have curated a set of 4 GPT models to assist in processing and predicting stock price movements.
The Challenge
After overcoming some challenges within my system, I was preparing to conduct a full-scale test when I encountered an obstacle in the form of an error message: "Error creating fine-tuning job: 429 This fine-tune request has been rate-limited. Your organisation has reached the maximum of 16 fine-tuning requests per day for the model ‘gpt-4o-mini-2024-07-18’."
Seeking Clarity
This limitation has hindered my ability to proceed with running tests on my entire dataset. Uncertainty looms as I seek information on potential monthly rate limits for fine-tuning, alongside exploring avenues to expand these limits temporarily for research purposes. I have initiated contact via email with [email protected] for clarification, yet I also wish to tap into the broader developer community for insights and guidance.
Community Support
If anyone has insights on monthly fine-tuning job caps for the model in question or has successfully requested limit increases for short-term purposes, your input would be invaluable. I am open to sharing specific code segments or additional details privately for a more in-depth discussion. Given the preliminary nature of my research work, ensuring ethical boundaries and confidentiality is paramount.
Expert Recommendations
For those facing similar challenges, it is essential to understand that limits are determined by Tier levels, with detailed guidelines accessible here. While daily limits pose a current hurdle, exploring options to elevate Tier status through account funding is a proactive step. Additionally, exceptions to limits can be petitioned, albeit with a potential waiting period for approval.
Exploring Solutions
As discussions unfold around exceeding daily caps and potential monthly constraints, the community's collective expertise can shed light on workarounds or alternative strategies to navigate these restrictions effectively. Every piece of advice contributes to a more robust understanding of fine-tuning processes and paves the way for innovative solutions.
Thank you for your anticipated support and valuable insights as I navigate through this challenge in fine-tuning GPT models for diverse applications within the financial domain.
