Cover image for OpenAI Takes Model Customization to the Next Level
Promptzone - Commumity
Promptzone - Commumity

Posted on

OpenAI Takes Model Customization to the Next Level

open ai fine tuning in api

Attention all developers and AI enthusiasts! OpenAI has just announced some exciting updates to their model customization offerings. Whether you're looking to fine-tune existing models or create custom-trained models from scratch, OpenAI has got you covered.

Fine-Tuning API Improvements

OpenAI's self-serve fine-tuning API for GPT-3.5 has been a game-changer since its launch in August 2023. Thousands of organizations have used it to train hundreds of thousands of models, tailoring them to specific tasks like code generation, text summarization, and personalized content creation.

But wait, there's more! OpenAI is now introducing new features to give developers even more control over their fine-tuning jobs:

  • Epoch-based Checkpoint Creation: Automatically produce full fine-tuned model checkpoints during each training epoch, reducing the need for retraining and preventing overfitting.
  • Comparative Playground: A side-by-side UI for comparing model quality and performance, allowing human evaluation of multiple models or fine-tune snapshots.
  • Third-party Integration: Support for integrating with platforms like Weights and Biases, letting you share detailed fine-tuning data with your existing stack.
  • Comprehensive Validation Metrics: Compute metrics like loss and accuracy over the entire validation dataset for better insight into model quality.
  • Hyperparameter Configuration: Configure available hyperparameters directly from the Dashboard, not just through the API or SDK.
  • Fine-Tuning Dashboard Improvements: Configure hyperparameters, view detailed training metrics, and rerun jobs from previous configurations.

Expanding Custom Models Program

OpenAI's Custom Model program, designed to train and optimize models for specific domains in partnership with their researchers, has been a resounding success. To further maximize performance, they're formally announcing their assisted fine-tuning offering.

Assisted fine-tuning is a collaborative effort with OpenAI's technical teams, leveraging techniques beyond the fine-tuning API, such as additional hyperparameters and parameter efficient fine-tuning (PEFT) methods at a larger scale. This is particularly helpful for organizations needing support with efficient training data pipelines, evaluation systems, and bespoke parameters and methods to maximize model performance for their use case or task.

For instance, SK Telecom, a South Korean telecommunications operator, worked with OpenAI to fine-tune GPT-4 for improved performance in telecom-related conversations in Korean. The result? A 35% increase in conversation summarization quality, a 33% increase in intent recognition accuracy, and a satisfaction score increase from 3.6 to 4.5 (out of 5) compared to GPT-4.

Custom-Trained Models

Sometimes, organizations need to train a purpose-built model from scratch that understands their business, industry, or domain. Fully custom-trained models imbue new knowledge from a specific domain by modifying key steps of the model training process using novel mid-training and post-training techniques.

OpenAI partnered with Harvey, an AI-native legal tool for attorneys, to create a custom-trained large language model for case law. After testing prompt engineering, RAG, and fine-tuning, they worked with OpenAI to add the depth of context needed – the equivalent of 10 billion tokens worth of data. The resulting model achieved an 83% increase in factual responses, and attorneys preferred its outputs 97% of the time over GPT-4.

The Future of Model Customization

OpenAI believes that in the future, most organizations will develop customized models personalized to their industry, business, or use case. With various techniques available to build custom models, organizations of all sizes can develop personalized models to realize more meaningful, specific impact from their AI implementations.

The key is to clearly scope the use case, design and implement evaluation systems, choose the right techniques, and be prepared to iterate over time for the model to reach optimal performance.

For most organizations, OpenAI's self-serve fine-tuning API can provide meaningful results quickly. For those needing to deeply fine-tune their models or imbue new, domain-specific knowledge, OpenAI's Custom Model programs are here to help.

Get Started Today

Visit OpenAI's fine-tuning API docs to start fine-tuning their models, or reach out to them for more information on how they can help customize models for your use case.

Stay ahead of the curve and unlock the full potential of AI with OpenAI's cutting-edge model customization offerings. The future of AI is here, and it's never been more exciting!

Check out this video for a quick overview of OpenAI's model customization offerings:

show case of fine tuning

Top comments (0)