OxMini-v2 / posttraining_log.csv
Shivam3002's picture
Publish measured OxMini v2 replay-SFT release
84bf611 verified
Raw History Blame Contribute Delete
859 Bytes
step,train_response_loss,train_replay_loss,val_response_loss,val_response_accuracy,learning_rate,elapsed_seconds
0,,,2.53867079,0.35189310,0.0,0.0
10,2.30709233,1.21929600,,,0.0000800000,52.843
20,1.95651857,1.32025892,2.16390808,0.40801782,0.0000788173,106.819
30,1.90871961,1.29350016,,,0.0000748286,159.404
40,1.77717760,1.33143444,2.06268531,0.43608018,0.0000683423,213.383
50,1.69500163,1.22210971,,,0.0000598840,266.575
60,1.62877946,1.29853139,2.04900321,0.44721604,0.0000501388,320.501
70,1.59936942,1.27816178,,,0.0000398963,372.473
80,1.54547579,1.32901660,2.01932963,0.44632517,0.0000299863,427.080
90,1.53246560,1.27868814,,,0.0000212116,479.207
100,1.49029202,1.25988015,2.00445652,0.45389755,0.0000142830,532.828
110,1.48356054,1.26788137,,,0.0000097620,586.724
120,1.47842176,1.31160543,2.00184938,0.45567929,0.0000080147,641.496