LoRA Training

Fine-tune models with LoRA adapters. Train custom adapters for Z-Image or Chronos-2 on your own data.

$50.00 per training job

Parameters

Parameter	Type	Required	Default	Description
model	string	required	—	Base model to fine-tuneOptions: zimage, chronos2
dataset_name	string	required	—	Name for the training dataset
train_values	float[][]	optional	—	Training data (for chronos2 — array of time series)
lora_r	int	optional	8	LoRA rank (higher = more capacity, more compute)
lora_alpha	int	optional	16	LoRA alpha scaling factor
learning_rate	float	optional	0.0001	Training learning rate
train_steps	int	optional	1000	Number of training steps
train_batch	int	optional	4	Training batch size

Example Request

curl -X POST https://cutedsl.cc/api/service \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -d '{
    "service": "lora_training",
    "model": "chronos2",
    "dataset_name": "my-timeseries",
    "train_values": [[1.0, 2.0, 3.0], [4.0, 5.0, 6.0]],
    "lora_r": 8,
    "train_steps": 1000
  }'

# Check training status:
curl https://cutedsl.cc/api/train/JOB_ID \
  -H "Authorization: Bearer YOUR_API_KEY"

Example Response

{
  "result": {
    "job_id": "train_abc123",
    "status": "queued",
    "model": "chronos2",
    "dataset_name": "my-timeseries",
    "estimated_time_minutes": 15
  },
  "credits_used": 50000,
  "credits_remain": 0,
  "usd_equivalent": 50
}

Playground

Connect your wallet to try the playground.

Connect Wallet

Back to all services