dataset,steps,ppl,entropy,self_bleu,seconds_per_sample lm1b,1,1.69,0.3001,0.9998,0.005960 lm1b,2,6962.13,4.7813,0.0083,0.011070 lm1b,4,1270.07,4.4491,0.0525,0.021469 lm1b,8,536.60,4.3759,0.0968,0.042353 lm1b,16,298.24,4.3469,0.1317,0.085267 lm1b,32,203.69,4.3330,0.1588,0.170720 lm1b,1024,121.73,4.3054,0.2049,5.460358 owt,1,14.53,1.0080,0.3997,0.009361 owt,2,28328.80,6.7991,0.0041,0.020205 owt,4,1553.66,6.0017,0.0580,0.041953 owt,8,514.24,5.8030,0.1009,0.085503 owt,16,227.83,5.6827,0.1318,0.172404 owt,32,121.06,5.5819,0.1548,0.346233 owt,1024,36.83,5.2608,0.1853,11.143945