LexLLMv0.0.0.x.10.8a / trainer_state.json
andrealexroom's picture
Upload folder using huggingface_hub
a36206c verified
{
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 0.450928611628209,
"eval_steps": 93,
"global_step": 651,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.0,
"learning_rate": 1e-05,
"loss": 1.2904,
"step": 1
},
{
"epoch": 0.06,
"learning_rate": 1e-05,
"loss": 1.2883,
"step": 93
},
{
"epoch": 0.06,
"eval_validation_privacy_sources_loss": 0.931429386138916,
"eval_validation_privacy_sources_runtime": 224.921,
"eval_validation_privacy_sources_samples_per_second": 1.258,
"eval_validation_privacy_sources_steps_per_second": 0.631,
"step": 93
},
{
"epoch": 0.06,
"eval_validation_agenda_digitale_loss": 1.4688478708267212,
"eval_validation_agenda_digitale_runtime": 201.5869,
"eval_validation_agenda_digitale_samples_per_second": 1.26,
"eval_validation_agenda_digitale_steps_per_second": 0.63,
"step": 93
},
{
"epoch": 0.06,
"eval_validation_leggepertutti_loss": 1.293728232383728,
"eval_validation_leggepertutti_runtime": 37.3205,
"eval_validation_leggepertutti_samples_per_second": 1.259,
"eval_validation_leggepertutti_steps_per_second": 0.643,
"step": 93
},
{
"epoch": 0.06,
"eval_validation_altalexprivacy_loss": 1.3008275032043457,
"eval_validation_altalexprivacy_runtime": 62.7339,
"eval_validation_altalexprivacy_samples_per_second": 1.259,
"eval_validation_altalexprivacy_steps_per_second": 0.638,
"step": 93
},
{
"epoch": 0.13,
"learning_rate": 1e-05,
"loss": 1.1632,
"step": 186
},
{
"epoch": 0.13,
"eval_validation_privacy_sources_loss": 0.8801236152648926,
"eval_validation_privacy_sources_runtime": 224.5622,
"eval_validation_privacy_sources_samples_per_second": 1.26,
"eval_validation_privacy_sources_steps_per_second": 0.632,
"step": 186
},
{
"epoch": 0.13,
"eval_validation_agenda_digitale_loss": 1.4332246780395508,
"eval_validation_agenda_digitale_runtime": 201.5945,
"eval_validation_agenda_digitale_samples_per_second": 1.26,
"eval_validation_agenda_digitale_steps_per_second": 0.63,
"step": 186
},
{
"epoch": 0.13,
"eval_validation_leggepertutti_loss": 1.2643470764160156,
"eval_validation_leggepertutti_runtime": 37.3205,
"eval_validation_leggepertutti_samples_per_second": 1.259,
"eval_validation_leggepertutti_steps_per_second": 0.643,
"step": 186
},
{
"epoch": 0.13,
"eval_validation_altalexprivacy_loss": 1.2626982927322388,
"eval_validation_altalexprivacy_runtime": 62.7289,
"eval_validation_altalexprivacy_samples_per_second": 1.259,
"eval_validation_altalexprivacy_steps_per_second": 0.638,
"step": 186
},
{
"epoch": 0.19,
"learning_rate": 1e-05,
"loss": 1.1339,
"step": 279
},
{
"epoch": 0.19,
"eval_validation_privacy_sources_loss": 0.8503767251968384,
"eval_validation_privacy_sources_runtime": 224.5704,
"eval_validation_privacy_sources_samples_per_second": 1.26,
"eval_validation_privacy_sources_steps_per_second": 0.632,
"step": 279
},
{
"epoch": 0.19,
"eval_validation_agenda_digitale_loss": 1.4141554832458496,
"eval_validation_agenda_digitale_runtime": 201.4979,
"eval_validation_agenda_digitale_samples_per_second": 1.261,
"eval_validation_agenda_digitale_steps_per_second": 0.63,
"step": 279
},
{
"epoch": 0.19,
"eval_validation_leggepertutti_loss": 1.2566713094711304,
"eval_validation_leggepertutti_runtime": 37.3026,
"eval_validation_leggepertutti_samples_per_second": 1.26,
"eval_validation_leggepertutti_steps_per_second": 0.643,
"step": 279
},
{
"epoch": 0.19,
"eval_validation_altalexprivacy_loss": 1.2368062734603882,
"eval_validation_altalexprivacy_runtime": 62.6933,
"eval_validation_altalexprivacy_samples_per_second": 1.26,
"eval_validation_altalexprivacy_steps_per_second": 0.638,
"step": 279
},
{
"epoch": 0.26,
"learning_rate": 1e-05,
"loss": 1.1007,
"step": 372
},
{
"epoch": 0.26,
"eval_validation_privacy_sources_loss": 0.8314403295516968,
"eval_validation_privacy_sources_runtime": 224.5386,
"eval_validation_privacy_sources_samples_per_second": 1.26,
"eval_validation_privacy_sources_steps_per_second": 0.632,
"step": 372
},
{
"epoch": 0.26,
"eval_validation_agenda_digitale_loss": 1.3996949195861816,
"eval_validation_agenda_digitale_runtime": 201.5206,
"eval_validation_agenda_digitale_samples_per_second": 1.26,
"eval_validation_agenda_digitale_steps_per_second": 0.63,
"step": 372
},
{
"epoch": 0.26,
"eval_validation_leggepertutti_loss": 1.2466578483581543,
"eval_validation_leggepertutti_runtime": 37.3026,
"eval_validation_leggepertutti_samples_per_second": 1.26,
"eval_validation_leggepertutti_steps_per_second": 0.643,
"step": 372
},
{
"epoch": 0.26,
"eval_validation_altalexprivacy_loss": 1.2185348272323608,
"eval_validation_altalexprivacy_runtime": 62.6898,
"eval_validation_altalexprivacy_samples_per_second": 1.26,
"eval_validation_altalexprivacy_steps_per_second": 0.638,
"step": 372
},
{
"epoch": 0.32,
"learning_rate": 1e-05,
"loss": 1.0747,
"step": 465
},
{
"epoch": 0.32,
"eval_validation_privacy_sources_loss": 0.8157272338867188,
"eval_validation_privacy_sources_runtime": 224.5077,
"eval_validation_privacy_sources_samples_per_second": 1.261,
"eval_validation_privacy_sources_steps_per_second": 0.632,
"step": 465
},
{
"epoch": 0.32,
"eval_validation_agenda_digitale_loss": 1.3890674114227295,
"eval_validation_agenda_digitale_runtime": 201.5006,
"eval_validation_agenda_digitale_samples_per_second": 1.261,
"eval_validation_agenda_digitale_steps_per_second": 0.63,
"step": 465
},
{
"epoch": 0.32,
"eval_validation_leggepertutti_loss": 1.2424800395965576,
"eval_validation_leggepertutti_runtime": 37.28,
"eval_validation_leggepertutti_samples_per_second": 1.261,
"eval_validation_leggepertutti_steps_per_second": 0.644,
"step": 465
},
{
"epoch": 0.32,
"eval_validation_altalexprivacy_loss": 1.203458547592163,
"eval_validation_altalexprivacy_runtime": 62.6722,
"eval_validation_altalexprivacy_samples_per_second": 1.261,
"eval_validation_altalexprivacy_steps_per_second": 0.638,
"step": 465
},
{
"epoch": 0.39,
"learning_rate": 1e-05,
"loss": 1.0703,
"step": 558
},
{
"epoch": 0.39,
"eval_validation_privacy_sources_loss": 0.8016490936279297,
"eval_validation_privacy_sources_runtime": 224.4789,
"eval_validation_privacy_sources_samples_per_second": 1.261,
"eval_validation_privacy_sources_steps_per_second": 0.633,
"step": 558
},
{
"epoch": 0.39,
"eval_validation_agenda_digitale_loss": 1.3786152601242065,
"eval_validation_agenda_digitale_runtime": 201.4702,
"eval_validation_agenda_digitale_samples_per_second": 1.261,
"eval_validation_agenda_digitale_steps_per_second": 0.63,
"step": 558
},
{
"epoch": 0.39,
"eval_validation_leggepertutti_loss": 1.23654043674469,
"eval_validation_leggepertutti_runtime": 37.2956,
"eval_validation_leggepertutti_samples_per_second": 1.26,
"eval_validation_leggepertutti_steps_per_second": 0.644,
"step": 558
},
{
"epoch": 0.39,
"eval_validation_altalexprivacy_loss": 1.1914178133010864,
"eval_validation_altalexprivacy_runtime": 62.6741,
"eval_validation_altalexprivacy_samples_per_second": 1.26,
"eval_validation_altalexprivacy_steps_per_second": 0.638,
"step": 558
},
{
"epoch": 0.45,
"learning_rate": 1e-05,
"loss": 1.0514,
"step": 651
},
{
"epoch": 0.45,
"eval_validation_privacy_sources_loss": 0.7911774516105652,
"eval_validation_privacy_sources_runtime": 224.3988,
"eval_validation_privacy_sources_samples_per_second": 1.261,
"eval_validation_privacy_sources_steps_per_second": 0.633,
"step": 651
},
{
"epoch": 0.45,
"eval_validation_agenda_digitale_loss": 1.370524525642395,
"eval_validation_agenda_digitale_runtime": 201.3956,
"eval_validation_agenda_digitale_samples_per_second": 1.261,
"eval_validation_agenda_digitale_steps_per_second": 0.631,
"step": 651
},
{
"epoch": 0.45,
"eval_validation_leggepertutti_loss": 1.2246781587600708,
"eval_validation_leggepertutti_runtime": 37.2851,
"eval_validation_leggepertutti_samples_per_second": 1.261,
"eval_validation_leggepertutti_steps_per_second": 0.644,
"step": 651
},
{
"epoch": 0.45,
"eval_validation_altalexprivacy_loss": 1.1746492385864258,
"eval_validation_altalexprivacy_runtime": 62.6484,
"eval_validation_altalexprivacy_samples_per_second": 1.261,
"eval_validation_altalexprivacy_steps_per_second": 0.638,
"step": 651
}
],
"logging_steps": 93,
"max_steps": 2886,
"num_input_tokens_seen": 0,
"num_train_epochs": 2,
"save_steps": 93,
"total_flos": 7.280849934083949e+18,
"train_batch_size": 1,
"trial_name": null,
"trial_params": null
}