d_hidden: 64
output_dim: 10
num_layers: 12
sigma: 0.03
batch_size: 256
num_steps: 5000
lr_fb: 0.001
lam: 0.1
K: 8
ema_momentum: 0.995
sigma_bridge: 0.03
eval_every: 200
gpu: 1
output_dir: results/toy_lq