NanoColibri / training_state.json
vovaRL's picture
step 20000 | vovaRL | 5242880000 tok
bfe502a verified
Raw
History Blame Contribute Delete
1.87 kB
{
"step": 20000,
"global_tokens": 5242880000,
"contributor": "vovaRL",
"cloud": "home",
"base_model": "scratch:colibri-nano",
"wall_clock_s": 24702.277401924133,
"history_tail": [
{
"step": 19960,
"t": 1785242935.469,
"ntp": 2.2264950275421143,
"zloss": 0.28825539350509644,
"lr": 3.000328985477171e-05,
"grad_norm": 0.28515625,
"ppl": 9.267327361978355,
"balance_max_imbalance": 0.1207275390625,
"balance_dead_frac": 0.0,
"tokens": 5232394240
},
{
"step": 19970,
"t": 1785243058.314,
"ntp": 2.558842420578003,
"zloss": 0.3002420961856842,
"lr": 3.000185054659739e-05,
"grad_norm": 0.302734375,
"ppl": 12.920851743202324,
"balance_max_imbalance": 0.06561279296875,
"balance_dead_frac": 0.0,
"tokens": 5235015680
},
{
"step": 19980,
"t": 1785243181.299,
"ntp": 2.413142442703247,
"zloss": 0.2963441014289856,
"lr": 3.0000822466198297e-05,
"grad_norm": 0.2734375,
"ppl": 11.169004011286283,
"balance_max_imbalance": 0.09515380859375,
"balance_dead_frac": 0.0,
"tokens": 5237637120
},
{
"step": 19990,
"t": 1785243304.446,
"ntp": 2.3685061931610107,
"zloss": 0.292042076587677,
"lr": 3.000020561670616e-05,
"grad_norm": 0.291015625,
"ppl": 10.681424375812318,
"balance_max_imbalance": 0.2261962890625,
"balance_dead_frac": 0.0,
"tokens": 5240258560
},
{
"step": 20000,
"t": 1785243427.437,
"ntp": 2.834693670272827,
"zloss": 0.2698419988155365,
"lr": 2.9999999999999997e-05,
"grad_norm": 0.28125,
"ppl": 17.025184181485457,
"balance_max_imbalance": 0.1982421875,
"balance_dead_frac": 0.0,
"tokens": 5242880000
}
]
}