step,train_loss,val_loss,lr,grad_norm,dt_ms,tok_per_sec,timer_data,timer_fwd,timer_loss,timer_bwd,timer_clip,timer_optim,mem_gpu_mb
0,11.160890,-1.000000,0.000007,117.933411,1168.004272,3506.836487,7.127712,962.354614,3.573728,49.161983,9.770976,113.502495,24404.937500