Faster Boot Times for Replicate's Fine-tuned Models

Replicate has significantly improved the cold boot times for fine-tuned models, with boot times now under one second.

The update applies to various fine-tuned language models such as Llama 2 and image models like SDXL.

This performance enhancement is currently only available for new models fine-tuned from the date of the announcement.

Replicate is working on further improvements to cold boot times that will be applicable to all models.

Users are encouraged to visit guides for fine-tuning Llama 2 and SDXL to take advantage of the new feature.