Skip to main content

​
What this is

Once a trainer is ready, create a client from the service endpoint and use it for train steps and checkpoint operations.

​
End-to-end examples

​
Create training client from RLOR endpoint

import tinker

service = tinker.ServiceClient(base_url="https://<direct-route-handle>", api_key="<FIREWORKS_API_KEY>")
training_client = service.create_lora_training_client(base_model="accounts/fireworks/models/qwen3-8b", rank=0)

​
Operational guidance

  • Service-mode trainer jobs currently support full-parameter tuning only. Use rank=0 (and lora_rank=0 during job creation).
  • Retry client creation while trainer status is still warming up.