model: Qwen/Qwen3-0.6B
model_class: llm


dist_config:
  name: fsdp2

optim_config:
  name: muon
  wd: 0.1
  momentum: 0.95
  nesterov: true
  ns_steps: 5
  adamw_betas: [0.9, 0.95]
  adamw_eps: 1.0e-8

### data
train_dataset: data/v1_sft_demo.yaml

### training
output_dir: outputs/test_muon
micro_batch_size: 1
cutoff_len: 2048
learning_rate: 1.0e-5
max_steps: 10

### sample
sample_backend: hf
max_new_tokens: 128