reasoning_hp_ablations_bsz64 / train_results.json
sedrickkeh's picture
End of training
b35a6e2 verified
{
"epoch": 3.0,
"total_flos": 4620219213545472.0,
"train_loss": 0.42465074877713965,
"train_runtime": 123675.358,
"train_samples_per_second": 2.764,
"train_steps_per_second": 0.043
}