Fastino Fine-Tune

Run managed config training for the GLiNER encoders fastino/gliner2-base-v1 and fastino/GLiNER-2.5-Decide from an uploaded dataset. Config jobs use SFT only. Query GET /v1/base-models?supports_training=true before selecting a base model.

Config jobs don't accept decoder base models; POST /v1/training/jobs returns 400 for them. Train decoders step by step: create an interactive job with POST /v1/training/jobs/create-model, then call /forward-backward, /optim-step, /save-weights-for-sampler, /sample, and deploy a sampler checkpoint. Follow https://docs.fastino.ai/training/training-loop.

Authentication

Fastino API keys start with fast_sk_ (pio_sk_ keys were retired on the move to Fastino). Both header formats are accepted: