PyTorch
llama
alignment-handbook
Generated from Trainer
Mamba2InLlama_0_75 / all_results.json
Junxiong Wang
add models
2804acb
raw
history blame contribute delete
765 Bytes
{
"epoch": 1.0,
"eval_logits/chosen": 1.226016640663147,
"eval_logits/rejected": 1.1679025888442993,
"eval_logps/chosen": -449.40472412109375,
"eval_logps/rejected": -588.7138671875,
"eval_loss": 0.4695761799812317,
"eval_rewards/accuracies": 0.8089285492897034,
"eval_rewards/chosen": -1.546373724937439,
"eval_rewards/margins": 1.3221595287322998,
"eval_rewards/rejected": -2.8685333728790283,
"eval_runtime": 173.1226,
"eval_samples": 4461,
"eval_samples_per_second": 25.768,
"eval_steps_per_second": 0.404,
"total_flos": 0.0,
"train_loss": 0.5209795107882678,
"train_runtime": 14376.1789,
"train_samples": 133368,
"train_samples_per_second": 9.277,
"train_steps_per_second": 0.29
}