Hi @porton
- The “Sync” link under the “Model” setting is indeed designed to allow you to access your own fine-tuned models. However, we have a feature that might be of interest to you: Embeddings. It offers a powerful alternative to fine-tuning, with potential for even higher performance, easier and cheaper.
- With regards to cost, using a fine-tuned model can indeed potentially help reduce the computational resources necessary to process each chatbot interaction. However, the cost savings can be variable depending on the specifics of the model and the nature of the interactions.
- Fine-tuning a model with the same short text that is used as context is not quite the same as just using it as context. When you fine-tune a model, you’re essentially training it to better understand or predict a certain type of data or specific tasks. Using the text for context, on the other hand, just gives the model some initial information to base its responses on but does not inherently change its understanding or prediction capabilities.
With fine-tuning, the model learns and adapts to the text, leading to more accurate responses in future interactions. This could be beneficial if you are dealing with a specific domain or type of conversation. However, it does not replace the need for providing context to the model for each interaction.
Now, about Embeddings: you can access them from the plugin menu under “Embeddings”. This could potentially provide even better performance for your chatbot, especially in understanding and responding to complex or nuanced queries.
You can watch our tutorial here:
https://www.youtube.com/watch?v=t3UQQ5-oNso
https://www.youtube.com/watch?v=NPMLGwFQYrY
