Nikita Pavlichenko
Calc loss only on prompts, add special tokens, remove grouping
77dd825
raw
history blame
195 Bytes
{
"epoch": 2.0,
"train_loss": 2.1184611846914603,
"train_runtime": 4663.4223,
"train_samples": 43003,
"train_samples_per_second": 18.443,
"train_steps_per_second": 2.306
}