Together AI enables fine-tuning of Hugging Face Hub models via API, including private repos

AI Impact Summary

Together AI is extending its platform to fine-tune any compatible Hugging Face Hub model using a two-model setup: a base training template (model) that configures resources and pipelines, and a custom HF model (from_hf_model) that gets fine-tuned. The integration can operate on private repositories by supplying HF tokens and can push the final finely-tuned model back to the Hub, enabling closed-loop improvements and team sharing. This dramatically lowers the infra and DevOps overhead for domain adaptation (healthcare, coding, multilingual tasks) and can reduce time-to-value and costs, but it introduces governance considerations around data provenance and secure token management when training on shared infrastructure.