new data

logs/main_log.txt

@@ -76874,3 +76874,67 @@ time (ms)
 time (ms)
  iteration      518/  292968 | consumed samples:      1060864 | consumed tokens:     79101952 | elapsed time per iteration (ms): 111964.5 | learning rate: 2.829E-05 | global batch size:  2048 | lm loss: 5.635888E+00 | loss scale: 8192.0 | grad norm: 12731.825 | num zeros: 0.0 | curriculum seqlen:    88 | number of skipped iterations:   0 | number of nan iterations:   0 |
 time (ms)
+ iteration      519/  292968 | consumed samples:      1062912 | consumed tokens:     79282176 | elapsed time per iteration (ms): 112055.2 | learning rate: 2.834E-05 | global batch size:  2048 | lm loss: 5.601050E+00 | loss scale: 8192.0 | grad norm: 11635.834 | num zeros: 0.0 | curriculum seqlen:    88 | number of skipped iterations:   0 | number of nan iterations:   0 |
+time (ms)
+ iteration      520/  292968 | consumed samples:      1064960 | consumed tokens:     79462400 | elapsed time per iteration (ms): 112418.5 | learning rate: 2.840E-05 | global batch size:  2048 | lm loss: 5.645939E+00 | loss scale: 8192.0 | grad norm: 17715.201 | num zeros: 0.0 | curriculum seqlen:    88 | number of skipped iterations:   0 | number of nan iterations:   0 |
+time (ms)
+ iteration      521/  292968 | consumed samples:      1067008 | consumed tokens:     79642624 | elapsed time per iteration (ms): 111558.5 | learning rate: 2.845E-05 | global batch size:  2048 | lm loss: 5.586247E+00 | loss scale: 8192.0 | grad norm: 9433.316 | num zeros: 0.0 | curriculum seqlen:    88 | number of skipped iterations:   0 | number of nan iterations:   0 |
+time (ms)
+ iteration      522/  292968 | consumed samples:      1069056 | consumed tokens:     79822848 | elapsed time per iteration (ms): 113098.8 | learning rate: 2.851E-05 | global batch size:  2048 | lm loss: 5.607241E+00 | loss scale: 8192.0 | grad norm: 11954.691 | num zeros: 0.0 | curriculum seqlen:    88 | number of skipped iterations:   0 | number of nan iterations:   0 |
+time (ms)
+ iteration      523/  292968 | consumed samples:      1071104 | consumed tokens:     80003072 | elapsed time per iteration (ms): 112106.8 | learning rate: 2.856E-05 | global batch size:  2048 | lm loss: 5.652853E+00 | loss scale: 8192.0 | grad norm: 16648.802 | num zeros: 0.0 | curriculum seqlen:    88 | number of skipped iterations:   0 | number of nan iterations:   0 |
+time (ms)
+ iteration      524/  292968 | consumed samples:      1073152 | consumed tokens:     80183296 | elapsed time per iteration (ms): 112809.7 | learning rate: 2.862E-05 | global batch size:  2048 | lm loss: 5.599886E+00 | loss scale: 8192.0 | grad norm: 9193.022 | num zeros: 0.0 | curriculum seqlen:    88 | number of skipped iterations:   0 | number of nan iterations:   0 |
+time (ms)
+ iteration      525/  292968 | consumed samples:      1075200 | consumed tokens:     80363520 | elapsed time per iteration (ms): 114026.4 | learning rate: 2.867E-05 | global batch size:  2048 | lm loss: 5.635831E+00 | loss scale: 8192.0 | grad norm: 22370.033 | num zeros: 0.0 | curriculum seqlen:    88 | number of skipped iterations:   0 | number of nan iterations:   0 |
+time (ms)
+ iteration      526/  292968 | consumed samples:      1077248 | consumed tokens:     80543744 | elapsed time per iteration (ms): 112873.4 | learning rate: 2.873E-05 | global batch size:  2048 | lm loss: 5.630721E+00 | loss scale: 8192.0 | grad norm: 11212.895 | num zeros: 0.0 | curriculum seqlen:    88 | number of skipped iterations:   0 | number of nan iterations:   0 |
+time (ms)
+ iteration      527/  292968 | consumed samples:      1079296 | consumed tokens:     80723968 | elapsed time per iteration (ms): 112562.2 | learning rate: 2.878E-05 | global batch size:  2048 | lm loss: 5.617833E+00 | loss scale: 8192.0 | grad norm: 16194.164 | num zeros: 0.0 | curriculum seqlen:    88 | number of skipped iterations:   0 | number of nan iterations:   0 |
+time (ms)
+ iteration      528/  292968 | consumed samples:      1081344 | consumed tokens:     80904192 | elapsed time per iteration (ms): 112871.5 | learning rate: 2.884E-05 | global batch size:  2048 | lm loss: 5.614437E+00 | loss scale: 8192.0 | grad norm: 13321.010 | num zeros: 0.0 | curriculum seqlen:    88 | number of skipped iterations:   0 | number of nan iterations:   0 |
+time (ms)
+ iteration      529/  292968 | consumed samples:      1083392 | consumed tokens:     81084416 | elapsed time per iteration (ms): 112230.3 | learning rate: 2.889E-05 | global batch size:  2048 | lm loss: 5.596371E+00 | loss scale: 8192.0 | grad norm: 9818.933 | num zeros: 0.0 | curriculum seqlen:    88 | number of skipped iterations:   0 | number of nan iterations:   0 |
+time (ms)
+ iteration      530/  292968 | consumed samples:      1085440 | consumed tokens:     81264640 | elapsed time per iteration (ms): 111781.8 | learning rate: 2.895E-05 | global batch size:  2048 | lm loss: 5.628756E+00 | loss scale: 8192.0 | grad norm: 15970.761 | num zeros: 0.0 | curriculum seqlen:    88 | number of skipped iterations:   0 | number of nan iterations:   0 |
+time (ms)
+ iteration      531/  292968 | consumed samples:      1087488 | consumed tokens:     81444864 | elapsed time per iteration (ms): 112070.8 | learning rate: 2.900E-05 | global batch size:  2048 | lm loss: 5.574606E+00 | loss scale: 8192.0 | grad norm: 12453.852 | num zeros: 0.0 | curriculum seqlen:    88 | number of skipped iterations:   0 | number of nan iterations:   0 |
+time (ms)
+ iteration      532/  292968 | consumed samples:      1089536 | consumed tokens:     81625088 | elapsed time per iteration (ms): 111479.9 | learning rate: 2.905E-05 | global batch size:  2048 | lm loss: 5.553162E+00 | loss scale: 8192.0 | grad norm: 12601.321 | num zeros: 0.0 | curriculum seqlen:    88 | number of skipped iterations:   0 | number of nan iterations:   0 |
+time (ms)
+ iteration      533/  292968 | consumed samples:      1091584 | consumed tokens:     81805312 | elapsed time per iteration (ms): 111390.0 | learning rate: 2.911E-05 | global batch size:  2048 | lm loss: 5.609733E+00 | loss scale: 8192.0 | grad norm: 13511.849 | num zeros: 0.0 | curriculum seqlen:    88 | number of skipped iterations:   0 | number of nan iterations:   0 |
+time (ms)
+ iteration      534/  292968 | consumed samples:      1093632 | consumed tokens:     81985536 | elapsed time per iteration (ms): 112213.7 | learning rate: 2.916E-05 | global batch size:  2048 | lm loss: 5.583689E+00 | loss scale: 8192.0 | grad norm: 11190.455 | num zeros: 0.0 | curriculum seqlen:    88 | number of skipped iterations:   0 | number of nan iterations:   0 |
+time (ms)
+ iteration      535/  292968 | consumed samples:      1095680 | consumed tokens:     82165760 | elapsed time per iteration (ms): 112993.1 | learning rate: 2.922E-05 | global batch size:  2048 | lm loss: 5.653582E+00 | loss scale: 8192.0 | grad norm: 20818.658 | num zeros: 0.0 | curriculum seqlen:    88 | number of skipped iterations:   0 | number of nan iterations:   0 |
+time (ms)
+ iteration      536/  292968 | consumed samples:      1097728 | consumed tokens:     82345984 | elapsed time per iteration (ms): 112307.7 | learning rate: 2.927E-05 | global batch size:  2048 | lm loss: 5.611212E+00 | loss scale: 8192.0 | grad norm: 10362.696 | num zeros: 0.0 | curriculum seqlen:    88 | number of skipped iterations:   0 | number of nan iterations:   0 |
+time (ms)
+ iteration      537/  292968 | consumed samples:      1099776 | consumed tokens:     82526208 | elapsed time per iteration (ms): 112970.5 | learning rate: 2.933E-05 | global batch size:  2048 | lm loss: 5.618240E+00 | loss scale: 8192.0 | grad norm: 14839.821 | num zeros: 0.0 | curriculum seqlen:    88 | number of skipped iterations:   0 | number of nan iterations:   0 |
+time (ms)
+ iteration      538/  292968 | consumed samples:      1101824 | consumed tokens:     82706432 | elapsed time per iteration (ms): 113120.5 | learning rate: 2.938E-05 | global batch size:  2048 | lm loss: 5.594517E+00 | loss scale: 8192.0 | grad norm: 13605.480 | num zeros: 0.0 | curriculum seqlen:    88 | number of skipped iterations:   0 | number of nan iterations:   0 |
+time (ms)
+ iteration      539/  292968 | consumed samples:      1103872 | consumed tokens:     82886656 | elapsed time per iteration (ms): 112476.6 | learning rate: 2.944E-05 | global batch size:  2048 | lm loss: 5.556248E+00 | loss scale: 8192.0 | grad norm: 13800.093 | num zeros: 0.0 | curriculum seqlen:    88 | number of skipped iterations:   0 | number of nan iterations:   0 |
+time (ms)
+ iteration      540/  292968 | consumed samples:      1105920 | consumed tokens:     83066880 | elapsed time per iteration (ms): 114182.8 | learning rate: 2.949E-05 | global batch size:  2048 | lm loss: 5.591393E+00 | loss scale: 8192.0 | grad norm: 10588.037 | num zeros: 0.0 | curriculum seqlen:    88 | number of skipped iterations:   0 | number of nan iterations:   0 |
+time (ms)
+ iteration      541/  292968 | consumed samples:      1107968 | consumed tokens:     83247104 | elapsed time per iteration (ms): 110876.2 | learning rate: 2.955E-05 | global batch size:  2048 | lm loss: 5.556509E+00 | loss scale: 8192.0 | grad norm: 13801.950 | num zeros: 0.0 | curriculum seqlen:    88 | number of skipped iterations:   0 | number of nan iterations:   0 |
+time (ms)
+ iteration      542/  292968 | consumed samples:      1110016 | consumed tokens:     83427328 | elapsed time per iteration (ms): 111658.7 | learning rate: 2.960E-05 | global batch size:  2048 | lm loss: 5.569237E+00 | loss scale: 8192.0 | grad norm: 14005.832 | num zeros: 0.0 | curriculum seqlen:    88 | number of skipped iterations:   0 | number of nan iterations:   0 |
+time (ms)
+ iteration      543/  292968 | consumed samples:      1112064 | consumed tokens:     83607552 | elapsed time per iteration (ms): 112214.1 | learning rate: 2.966E-05 | global batch size:  2048 | lm loss: 5.546272E+00 | loss scale: 8192.0 | grad norm: 11650.584 | num zeros: 0.0 | curriculum seqlen:    88 | number of skipped iterations:   0 | number of nan iterations:   0 |
+time (ms)
+ iteration      544/  292968 | consumed samples:      1114112 | consumed tokens:     83787776 | elapsed time per iteration (ms): 113179.9 | learning rate: 2.971E-05 | global batch size:  2048 | lm loss: 5.549253E+00 | loss scale: 8192.0 | grad norm: 13630.378 | num zeros: 0.0 | curriculum seqlen:    88 | number of skipped iterations:   0 | number of nan iterations:   0 |
+time (ms)
+ iteration      545/  292968 | consumed samples:      1116160 | consumed tokens:     83968000 | elapsed time per iteration (ms): 112602.2 | learning rate: 2.976E-05 | global batch size:  2048 | lm loss: 5.533734E+00 | loss scale: 8192.0 | grad norm: 10491.189 | num zeros: 0.0 | curriculum seqlen:    88 | number of skipped iterations:   0 | number of nan iterations:   0 |
+time (ms)
+ iteration      546/  292968 | consumed samples:      1118208 | consumed tokens:     84148224 | elapsed time per iteration (ms): 112024.9 | learning rate: 2.982E-05 | global batch size:  2048 | lm loss: 5.555665E+00 | loss scale: 8192.0 | grad norm: 14130.965 | num zeros: 0.0 | curriculum seqlen:    88 | number of skipped iterations:   0 | number of nan iterations:   0 |
+time (ms)
+ iteration      547/  292968 | consumed samples:      1120256 | consumed tokens:     84328448 | elapsed time per iteration (ms): 112655.1 | learning rate: 2.987E-05 | global batch size:  2048 | lm loss: 5.551611E+00 | loss scale: 8192.0 | grad norm: 12855.412 | num zeros: 0.0 | curriculum seqlen:    88 | number of skipped iterations:   0 | number of nan iterations:   0 |
+time (ms)
+ iteration      548/  292968 | consumed samples:      1122304 | consumed tokens:     84508672 | elapsed time per iteration (ms): 111576.2 | learning rate: 2.993E-05 | global batch size:  2048 | lm loss: 5.609882E+00 | loss scale: 8192.0 | grad norm: 15275.244 | num zeros: 0.0 | curriculum seqlen:    88 | number of skipped iterations:   0 | number of nan iterations:   0 |
+time (ms)
+ iteration      549/  292968 | consumed samples:      1124352 | consumed tokens:     84688896 | elapsed time per iteration (ms): 112577.9 | learning rate: 2.998E-05 | global batch size:  2048 | lm loss: 5.596916E+00 | loss scale: 8192.0 | grad norm: 13652.963 | num zeros: 0.0 | curriculum seqlen:    88 | number of skipped iterations:   0 | number of nan iterations:   0 |
+time (ms)
+ iteration      550/  292968 | consumed samples:      1126400 | consumed tokens:     84869120 | elapsed time per iteration (ms): 112519.9 | learning rate: 3.004E-05 | global batch size:  2048 | lm loss: 5.550436E+00 | loss scale: 8192.0 | grad norm: 10479.605 | num zeros: 0.0 | curriculum seqlen:    88 | number of skipped iterations:   0 | number of nan iterations:   0 |
+time (ms)