HyperWrite has recently launched their new open-source tool ‘gpt-oracle-trainer’, which has the ability to fine-tune, use embeddings, and accurately answer questions about a product or service.
The experimental tool is a chatbot that aims to simplify the process of accurately answering queries about products or services with a simple text prompt.
In a recent tweet from Matt Shumer, HyperWrite CEO, he officially introduced the new chatbot with a short demonstration of how it works.
According to Matt, the tool is a constrained agent that chains together GPT calls that work together creating a good dataset for users.
Furthermore, the release enumerated its features namely;
- Data Generation - the tool generates a question-and-answer dataset based on a user’s service description, documentation, temperature, and number of examples indicated.
- Model Training - the tool trains the model using the generated dataset.
- Model Testing - the trained models can be tested using a custom prompt.
Users can simply paste in their product or service’s documentation and the system will automatically generate a dataset in the correct format, train the model, and come up with relevant questions and answers.
In a previous report, HyperWrite has also launched an open-source trainer called ‘gpt-llm-trainer’ where it simplifies complexities of AI model training, similar to this new release.
HyperWrite continues to make AI-training easier for users that even simple text prompts would help users refrain from doing complex tasks. The good thing is, these tools that simplify tasks can create great datasets and accurate answers – potentially producing better results.