ferdinand.mom
add metrics + profiler.csv for 16 GPUS
b79f60a
raw
history blame
2.25 kB
iteration,consumed_tokens,elapsed_time_per_iteration_ms,tokens_per_sec,tokens_per_sec_per_gpu,global_batch_size,lm_loss,lr,model_tflops_per_gpu,hardware_tflops_per_gpu,grad_norm,memory_usage_MiB,peak_allocated_MiB,peak_reserved_MiB
1,4190000.0000000005,18900.0,222000.0,13900.0,1020.0,11.3,0.0001,126.0,126.0,33.1,7252.46,25075.44,27568.0
2,8390000.0,9350.0,449000.0,28000.0,1020.0,11.3,9.53e-05,254.0,254.0,33.3,7252.46,25075.44,27568.0
3,12600000.0,9080.0,462000.0,28900.0,1020.0,16.0,9.05e-05,262.0,262.0,249.0,7252.46,25075.44,27568.0
4,16800000.0,9640.0,435000.0,27200.0,1020.0,15.1,8.58e-05,247.0,247.0,41.6,7252.45,11617.76,27568.0
5,21000000.0,9500.0,441000.0,27600.0,1020.0,10.8,8.11e-05,250.0,250.0,25.9,7252.45,25075.44,27568.0
6,25200000.0,9730.0,431000.0,26900.0,1020.0,10.8,7.63e-05,244.0,244.0,18.9,7252.45,25075.44,27568.0
7,29400000.0,8900.0,471000.0,29500.0,1020.0,10.2,7.16e-05,267.0,267.0,7.97,7252.45,25075.44,27568.0
8,33600000.0,8860.0,473000.0,29600.0,1020.0,9.15,6.68e-05,268.0,268.0,6.46,7252.45,25075.44,27568.0
9,37700000.0,9080.0,462000.0,28900.0,1020.0,11.2,6.21e-05,262.0,262.0,59.7,7252.45,25075.44,27568.0
10,41900000.0,9130.0,459000.0,28700.0,1020.0,9.6,5.74e-05,260.0,260.0,44.2,7252.45,25075.44,27568.0
11,46100000.0,9390.0,447000.0,27900.0,1020.0,8.08,5.26e-05,253.0,253.0,8.69,7252.45,25075.44,27568.0
12,50300000.0,8910.0,471000.0,29400.0,1020.0,7.86,4.79e-05,267.0,267.0,5.1,7252.45,25075.44,27568.0
13,54500000.0,9060.0,463000.0,28900.0,1020.0,7.7,4.32e-05,263.0,263.0,4.73,7252.45,25075.44,27568.0
14,58700000.0,9030.0,464000.0,29000.0,1020.0,7.56,3.84e-05,263.0,263.0,5.09,7252.45,25075.44,27568.0
15,62900000.0,9030.0,464000.0,29000.0,1020.0,7.4,3.37e-05,263.0,263.0,5.16,7252.45,25075.44,27568.0
16,67099999.99999999,9270.0,453000.0,28300.0,1020.0,7.3,2.89e-05,257.0,257.0,5.15,7252.45,25075.44,27568.0
17,71300000.0,9240.0,454000.0,28400.0,1020.0,7.22,2.42e-05,258.0,258.0,5.14,7252.45,25075.44,27568.0
18,75500000.0,9120.0,460000.0,28800.0,1020.0,7.15,1.95e-05,261.0,261.0,5.04,7252.45,25075.44,27568.0
19,79700000.0,9110.0,460000.0,28800.0,1020.0,7.08,1.47e-05,261.0,261.0,3.86,7252.45,25075.44,27568.0
20,83900000.0,8990.0,467000.0,29200.0,1020.0,7.03,1e-05,265.0,265.0,2.94,,,