OpenAI models reject 'max_tokens' and require 'max_completion_tokens'. Also skip 'top_k' for OpenAI as it is unsupported. Detection is based on the provider field in the model catalog.
100 KiB
100 KiB