Files
Mistral-7B-Instruct-v0.3-hh…/lr_sweep_summary.json

52 lines
1.1 KiB
JSON
Raw Permalink Normal View History

{
"selection": {
"top_k_by_val_loss": 3,
"selected_learning_rate": 1e-05,
"selected_eval_loss": 0.5897703170776367,
"selected_eval_accuracy": null
},
"all_results": [
{
"learning_rate": 5e-07,
"eval_loss": 0.6795729398727417,
"eval_accuracy": null
},
{
"learning_rate": 1e-06,
"eval_loss": 0.6668933629989624,
"eval_accuracy": null
},
{
"learning_rate": 2e-06,
"eval_loss": 0.6433066129684448,
"eval_accuracy": null
},
{
"learning_rate": 5e-06,
"eval_loss": 0.608209490776062,
"eval_accuracy": null
},
{
"learning_rate": 1e-05,
"eval_loss": 0.5897703170776367,
"eval_accuracy": null
}
],
"top_by_val_loss": [
{
"learning_rate": 1e-05,
"eval_loss": 0.5897703170776367,
"eval_accuracy": null
},
{
"learning_rate": 5e-06,
"eval_loss": 0.608209490776062,
"eval_accuracy": null
},
{
"learning_rate": 2e-06,
"eval_loss": 0.6433066129684448,
"eval_accuracy": null
}
]
}