data-ladder: dl_smollm2_135m_dclm

data-ladder-runs/dl_smollm2_135m_dclm/metrics.csv

@@ -0,0 +1,31 @@
+step,tokens,timed_hours,gpu_hours,train_loss_ema,neutral_bpb,ownval_bpb,lr,wallclock_s
+17,8912896,0.001219,0.009751,10.2144,3.287973,2.943902,3.02e-04,42.2
+23,12058624,0.002312,0.018496,9.6149,3.079832,2.706418,4.08e-04,76.5
+30,15728640,0.003469,0.027755,8.9645,2.990359,2.582599,5.33e-04,111.1
+40,20971520,0.005189,0.041512,8.2669,2.878704,2.500581,7.10e-04,147.6
+54,28311552,0.007601,0.060806,7.5804,2.772781,2.382689,9.59e-04,186.8
+71,37224448,0.010513,0.084104,6.9775,2.631513,2.262019,1.26e-03,227.8
+95,49807360,0.014392,0.115134,6.4462,2.521288,2.152354,1.69e-03,272.2
+127,66584576,0.019670,0.157356,6.0225,2.490505,2.070549,2.25e-03,321.4
+169,88604672,0.026501,0.212009,5.6591,2.263582,1.910550,3.00e-03,376.3
+226,118489088,0.035584,0.284670,5.2133,2.098044,1.768919,3.00e-03,439.3
+301,157810688,0.047488,0.379907,4.7335,1.827166,1.612165,3.00e-03,512.3
+401,210239488,0.063099,0.504790,4.3194,1.618772,1.484862,3.00e-03,598.7
+535,280494080,0.083620,0.668957,4.0733,1.506554,1.403185,3.00e-03,702.8
+713,373817344,0.110824,0.886591,3.8878,1.436450,1.344268,3.00e-03,831.0
+951,498597888,0.146021,1.168169,3.7461,1.384255,1.299763,3.00e-03,987.9
+1269,665321472,0.192554,1.540433,3.6248,1.334495,1.260089,3.00e-03,1185.5
+1692,887095296,0.254014,2.032110,3.5518,1.301475,1.228359,3.00e-03,1437.0
+2256,1182793728,0.335296,2.682369,3.4774,1.268740,1.201851,3.00e-03,1759.7
+3009,1577582592,0.443570,3.548557,3.4011,1.244917,1.178775,3.00e-03,2179.7
+4012,2103443456,0.587623,4.700985,3.3539,1.225674,1.158683,3.00e-03,2728.5
+5350,2804940800,0.779754,6.238033,3.3034,1.208051,1.142357,3.00e-03,3450.5
+7135,3740794880,1.036047,8.288380,3.2611,1.189133,1.126976,3.00e-03,4403.5
+9514,4988076032,1.377602,11.020812,3.2302,1.172962,1.112073,3.00e-03,5663.5
+12687,6651641856,1.833177,14.665414,3.1868,1.163154,1.100147,3.00e-03,7333.8
+15227,7983333376,2.197868,17.582945,3.1238,1.141277,1.079524,1.52e-03,8677.0
+15565,8160542720,2.246391,17.971125,3.1245,1.134419,1.075549,1.21e-03,8882.1
+15904,8338276352,2.295060,18.360477,3.1169,1.132241,1.071580,9.10e-04,9087.6
+16242,8515485696,2.343580,18.748643,3.1028,1.127143,1.068065,6.07e-04,9292.4
+16581,8693219328,2.392249,19.137990,3.1079,1.125012,1.064920,3.04e-04,9498.0
+16919,8870428672,2.440784,19.526270,3.1005,1.122708,1.062973,8.96e-07,9703.1