[INFO] Training `lr` on 800 samples (%80.00 of 1,000 samples)...
