LiteLong_NExtLong_128K / train_results.json
jaber
upload initial model
32b2389
{
"epoch": 0.0005,
"num_input_tokens_seen": 8396996608,
"train_loss": 0.001497477203696877,
"train_runtime": 201.1036,
"train_samples_per_second": 318.244,
"train_steps_per_second": 4.973
}