Create a fine-tune training job
Render the GPU training DAG from an approved dataset job and persist it.
Authorizations
The access token received from the authorization server in the OAuth 2.0 flow.
Body
Start a fine-tune training run from an approved dataset job.
A COMPLETE fine-tune dataset job supplying the JSONL + base model.
Human-friendly job name. Defaults to '-<base_model>'.
256Parameter-efficient fine-tuning method.
lora, qlora Trainer hyperparameters (epochs, learning_rate, lora_r, lora_alpha, lora_dropout, max_seq_len, ...). Trainer-side defaults fill any gaps.
Override GPU count; clamped to the configured min/max.
1 <= x <= 64Trigger the Airflow DAG immediately after creation.
Response
Job created with rendered DAG.
Full state of a fine-tune training job.
Parameter-efficient fine-tuning method.
lora, qlora Lifecycle of a fine-tune training job (mirrors the DB enum).
ready, queued, running, complete, failed 
