OpenAI Chat Model Fine-Tuner

This node allows you to fine-tune an OpenAI Chat Model based on a conversation table.

The fine-tuning data needs to be in the following format and contain at least 10 conversations, each of which must contain at least one system message:

IDRoleContent1systemYou are a happy assistant that puts a positive spin on everything.1userI lost my tennis match today.1assistantIt's ok, it happens to everyone.2userI lost my book today.2assistantYou can read everything on ebooks these days!id_stringsystemYou are a happy assistant that puts a positive spin on everything.id_stringassistantYou're great!

For a training file with 100,000 tokens trained over 3 epochs, the expected cost would be ~$2.40 USD. For more information, visit OpenAI

Note : If you use the Credentials Configuration node and do not select the "Save password in configuration (weakly encrypted)" option for passing the API key, the Credentials Configuration node will need to be reconfigured upon reopening the workflow, as the credentials flow variable was not saved and will therefore not be available to downstream nodes.

Input ports

1. Type: org.knime.python3.nodes.PythonBinaryBlobFileStorePortObject
   OpenAI Chat Model

   Configured OpenAI Chat Model which supports fine-tuning.

2. Type: Table
   Fine-tuning Data

   The data should be presented across 3 columns:

   One column specifying a conversation ID, one representing the role of a message (system, assistant and user), and the third for the content of the message.

   The table has to include at least 10 conversations, each of which must contain at least one system message.

Output ports

1. Type: org.knime.python3.nodes.PythonBinaryBlobFileStorePortObject
   OpenAI Chat Model

   Configured fine-tuned OpenAI Chat Model connection.

2. Type: Table
   Fine-tuning Metrics

   Metrics to evaluate the fine-tuning performance. The values of the metrics are: 'train loss', 'train accuracy', 'valid loss', and 'valid mean token accuracy' for each step of training.