ferdinand.mom
add metrics + profiler.csv for 16 GPUS
b79f60a
raw
history blame
2.12 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
1,4190000.0000000005,81400.0,51500.0,3220.0,1020.0,11.1,0.0001,29.2,29.2,25.6,3274.15,3274.15,43028.0
2,8390000.0,30800.0,136000.0,8500.0,1020.0,11.1,9.53e-05,77.1,77.1,25.9,3274.15,3274.15,43028.0
3,12600000.0,28800.0,145000.0,9090.0,1020.0,9.9,9.05e-05,82.5,82.5,40.4,3274.15,3274.15,43028.0
4,16800000.0,30000.0,140000.0,8750.0,1020.0,11.9,8.58e-05,79.4,79.4,61.2,3274.15,42592.08,43028.0
5,21000000.0,31800.0,132000.0,8250.0,1020.0,9.05,8.11e-05,74.9,74.9,8.31,,,
6,25200000.0,30600.0,137000.0,8550.0,1020.0,8.85,7.63e-05,77.6,77.6,6.61,3274.15,42592.08,43028.0
7,29400000.0,146000.0,28700.0,1790.0,1020.0,8.37,7.16e-05,16.3,16.3,4.93,,,
8,33600000.0,29800.0,141000.0,8810.0,1020.0,7.97,6.68e-05,79.9,79.9,3.12,3274.15,42592.08,43028.0
9,37700000.0,31500.0,133000.0,8320.0,1020.0,7.83,6.21e-05,75.5,75.5,9.04,,,
10,41900000.0,31400.0,134000.0,8350.0,1020.0,7.62,5.74e-05,75.8,75.8,5.09,3274.15,42592.08,43028.0
11,46100000.0,30300.0,138000.0,8650.0,1020.0,7.47,5.26e-05,78.5,78.5,4.06,3274.15,42592.08,43028.0
12,50300000.0,32000.0,131000.0,8180.0,1020.0,7.34,4.79e-05,74.2,74.2,3.13,,,
13,54500000.0,32400.0,129000.0,8080.0,1020.0,7.23,4.32e-05,73.4,73.4,2.73,3274.15,42592.08,43028.0
14,58700000.0,30900.0,136000.0,8480.0,1020.0,7.14,3.84e-05,77.0,77.0,2.33,3274.15,42592.08,43028.0
15,62900000.0,30600.0,137000.0,8560.0,1020.0,7.06,3.37e-05,77.7,77.7,2.47,3274.15,42592.08,43028.0
16,67099999.99999999,32100.0,131000.0,8170.0,1020.0,6.98,2.89e-05,74.1,74.1,2.69,3274.15,42592.08,43028.0
17,71300000.0,31300.0,134000.0,8370.0,1020.0,6.9,2.42e-05,75.9,75.9,1.91,3274.15,42592.08,43028.0
18,75500000.0,29900.0,140000.0,8770.0,1020.0,6.84,1.95e-05,79.5,79.5,1.62,3274.15,42592.08,43028.0
19,79700000.0,31700.0,132000.0,8270.0,1020.0,6.8,1.47e-05,75.0,75.0,1.85,3274.15,42592.08,43028.0
20,83900000.0,30800.0,136000.0,8500.0,1020.0,6.77,1e-05,77.1,77.1,1.82,,,