LexLLMv0.0.0.x.10.10a / trainer_state.json
andrealexroom's picture
Upload folder using huggingface_hub
1c3f606 verified
{
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 0.21192502812506633,
"eval_steps": 78,
"global_step": 624,
"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": 1.2868,
"step": 1
},
{
"epoch": 0.03,
"learning_rate": 1e-05,
"loss": 1.1922,
"step": 78
},
{
"epoch": 0.03,
"eval_validation_privacy_sources_loss": 1.0002105236053467,
"eval_validation_privacy_sources_runtime": 224.8773,
"eval_validation_privacy_sources_samples_per_second": 1.258,
"eval_validation_privacy_sources_steps_per_second": 0.631,
"step": 78
},
{
"epoch": 0.03,
"eval_validation_agenda_digitale_loss": 1.506428837776184,
"eval_validation_agenda_digitale_runtime": 201.6135,
"eval_validation_agenda_digitale_samples_per_second": 1.26,
"eval_validation_agenda_digitale_steps_per_second": 0.63,
"step": 78
},
{
"epoch": 0.03,
"eval_validation_leggepertutti_loss": 1.3169218301773071,
"eval_validation_leggepertutti_runtime": 37.3198,
"eval_validation_leggepertutti_samples_per_second": 1.259,
"eval_validation_leggepertutti_steps_per_second": 0.643,
"step": 78
},
{
"epoch": 0.03,
"eval_validation_altalexprivacy_loss": 1.3375698328018188,
"eval_validation_altalexprivacy_runtime": 62.731,
"eval_validation_altalexprivacy_samples_per_second": 1.259,
"eval_validation_altalexprivacy_steps_per_second": 0.638,
"step": 78
},
{
"epoch": 0.05,
"learning_rate": 1e-05,
"loss": 1.0426,
"step": 156
},
{
"epoch": 0.05,
"eval_validation_privacy_sources_loss": 0.9441261887550354,
"eval_validation_privacy_sources_runtime": 224.5852,
"eval_validation_privacy_sources_samples_per_second": 1.26,
"eval_validation_privacy_sources_steps_per_second": 0.632,
"step": 156
},
{
"epoch": 0.05,
"eval_validation_agenda_digitale_loss": 1.4715445041656494,
"eval_validation_agenda_digitale_runtime": 201.5442,
"eval_validation_agenda_digitale_samples_per_second": 1.26,
"eval_validation_agenda_digitale_steps_per_second": 0.63,
"step": 156
},
{
"epoch": 0.05,
"eval_validation_leggepertutti_loss": 1.290155291557312,
"eval_validation_leggepertutti_runtime": 37.3113,
"eval_validation_leggepertutti_samples_per_second": 1.26,
"eval_validation_leggepertutti_steps_per_second": 0.643,
"step": 156
},
{
"epoch": 0.05,
"eval_validation_altalexprivacy_loss": 1.2974395751953125,
"eval_validation_altalexprivacy_runtime": 62.7333,
"eval_validation_altalexprivacy_samples_per_second": 1.259,
"eval_validation_altalexprivacy_steps_per_second": 0.638,
"step": 156
},
{
"epoch": 0.08,
"learning_rate": 1e-05,
"loss": 0.9843,
"step": 234
},
{
"epoch": 0.08,
"eval_validation_privacy_sources_loss": 0.9115681648254395,
"eval_validation_privacy_sources_runtime": 224.5909,
"eval_validation_privacy_sources_samples_per_second": 1.26,
"eval_validation_privacy_sources_steps_per_second": 0.632,
"step": 234
},
{
"epoch": 0.08,
"eval_validation_agenda_digitale_loss": 1.454103946685791,
"eval_validation_agenda_digitale_runtime": 201.5826,
"eval_validation_agenda_digitale_samples_per_second": 1.26,
"eval_validation_agenda_digitale_steps_per_second": 0.63,
"step": 234
},
{
"epoch": 0.08,
"eval_validation_leggepertutti_loss": 1.2771508693695068,
"eval_validation_leggepertutti_runtime": 37.3084,
"eval_validation_leggepertutti_samples_per_second": 1.26,
"eval_validation_leggepertutti_steps_per_second": 0.643,
"step": 234
},
{
"epoch": 0.08,
"eval_validation_altalexprivacy_loss": 1.2786180973052979,
"eval_validation_altalexprivacy_runtime": 62.71,
"eval_validation_altalexprivacy_samples_per_second": 1.26,
"eval_validation_altalexprivacy_steps_per_second": 0.638,
"step": 234
},
{
"epoch": 0.11,
"learning_rate": 1e-05,
"loss": 0.9674,
"step": 312
},
{
"epoch": 0.11,
"eval_validation_privacy_sources_loss": 0.8925028443336487,
"eval_validation_privacy_sources_runtime": 224.5822,
"eval_validation_privacy_sources_samples_per_second": 1.26,
"eval_validation_privacy_sources_steps_per_second": 0.632,
"step": 312
},
{
"epoch": 0.11,
"eval_validation_agenda_digitale_loss": 1.442084550857544,
"eval_validation_agenda_digitale_runtime": 201.5693,
"eval_validation_agenda_digitale_samples_per_second": 1.26,
"eval_validation_agenda_digitale_steps_per_second": 0.63,
"step": 312
},
{
"epoch": 0.11,
"eval_validation_leggepertutti_loss": 1.2704201936721802,
"eval_validation_leggepertutti_runtime": 37.2901,
"eval_validation_leggepertutti_samples_per_second": 1.26,
"eval_validation_leggepertutti_steps_per_second": 0.644,
"step": 312
},
{
"epoch": 0.11,
"eval_validation_altalexprivacy_loss": 1.2658644914627075,
"eval_validation_altalexprivacy_runtime": 62.696,
"eval_validation_altalexprivacy_samples_per_second": 1.26,
"eval_validation_altalexprivacy_steps_per_second": 0.638,
"step": 312
},
{
"epoch": 0.13,
"learning_rate": 1e-05,
"loss": 0.9451,
"step": 390
},
{
"epoch": 0.13,
"eval_validation_privacy_sources_loss": 0.8763552904129028,
"eval_validation_privacy_sources_runtime": 224.7039,
"eval_validation_privacy_sources_samples_per_second": 1.259,
"eval_validation_privacy_sources_steps_per_second": 0.632,
"step": 390
},
{
"epoch": 0.13,
"eval_validation_agenda_digitale_loss": 1.4315780401229858,
"eval_validation_agenda_digitale_runtime": 201.6786,
"eval_validation_agenda_digitale_samples_per_second": 1.259,
"eval_validation_agenda_digitale_steps_per_second": 0.63,
"step": 390
},
{
"epoch": 0.13,
"eval_validation_leggepertutti_loss": 1.2630696296691895,
"eval_validation_leggepertutti_runtime": 37.333,
"eval_validation_leggepertutti_samples_per_second": 1.259,
"eval_validation_leggepertutti_steps_per_second": 0.643,
"step": 390
},
{
"epoch": 0.13,
"eval_validation_altalexprivacy_loss": 1.2540370225906372,
"eval_validation_altalexprivacy_runtime": 62.7795,
"eval_validation_altalexprivacy_samples_per_second": 1.258,
"eval_validation_altalexprivacy_steps_per_second": 0.637,
"step": 390
},
{
"epoch": 0.16,
"learning_rate": 1e-05,
"loss": 0.926,
"step": 468
},
{
"epoch": 0.16,
"eval_validation_privacy_sources_loss": 0.8631904721260071,
"eval_validation_privacy_sources_runtime": 224.6404,
"eval_validation_privacy_sources_samples_per_second": 1.26,
"eval_validation_privacy_sources_steps_per_second": 0.632,
"step": 468
},
{
"epoch": 0.16,
"eval_validation_agenda_digitale_loss": 1.4235539436340332,
"eval_validation_agenda_digitale_runtime": 201.6226,
"eval_validation_agenda_digitale_samples_per_second": 1.26,
"eval_validation_agenda_digitale_steps_per_second": 0.63,
"step": 468
},
{
"epoch": 0.16,
"eval_validation_leggepertutti_loss": 1.2557491064071655,
"eval_validation_leggepertutti_runtime": 37.3222,
"eval_validation_leggepertutti_samples_per_second": 1.259,
"eval_validation_leggepertutti_steps_per_second": 0.643,
"step": 468
},
{
"epoch": 0.16,
"eval_validation_altalexprivacy_loss": 1.244572639465332,
"eval_validation_altalexprivacy_runtime": 62.7205,
"eval_validation_altalexprivacy_samples_per_second": 1.26,
"eval_validation_altalexprivacy_steps_per_second": 0.638,
"step": 468
},
{
"epoch": 0.19,
"learning_rate": 1e-05,
"loss": 0.9164,
"step": 546
},
{
"epoch": 0.19,
"eval_validation_privacy_sources_loss": 0.8481507301330566,
"eval_validation_privacy_sources_runtime": 224.6424,
"eval_validation_privacy_sources_samples_per_second": 1.26,
"eval_validation_privacy_sources_steps_per_second": 0.632,
"step": 546
},
{
"epoch": 0.19,
"eval_validation_agenda_digitale_loss": 1.4142533540725708,
"eval_validation_agenda_digitale_runtime": 201.7172,
"eval_validation_agenda_digitale_samples_per_second": 1.259,
"eval_validation_agenda_digitale_steps_per_second": 0.63,
"step": 546
},
{
"epoch": 0.19,
"eval_validation_leggepertutti_loss": 1.2529613971710205,
"eval_validation_leggepertutti_runtime": 37.3269,
"eval_validation_leggepertutti_samples_per_second": 1.259,
"eval_validation_leggepertutti_steps_per_second": 0.643,
"step": 546
},
{
"epoch": 0.19,
"eval_validation_altalexprivacy_loss": 1.2327677011489868,
"eval_validation_altalexprivacy_runtime": 62.729,
"eval_validation_altalexprivacy_samples_per_second": 1.259,
"eval_validation_altalexprivacy_steps_per_second": 0.638,
"step": 546
},
{
"epoch": 0.21,
"learning_rate": 1e-05,
"loss": 0.9088,
"step": 624
},
{
"epoch": 0.21,
"eval_validation_privacy_sources_loss": 0.8413664698600769,
"eval_validation_privacy_sources_runtime": 224.6765,
"eval_validation_privacy_sources_samples_per_second": 1.26,
"eval_validation_privacy_sources_steps_per_second": 0.632,
"step": 624
},
{
"epoch": 0.21,
"eval_validation_agenda_digitale_loss": 1.4095916748046875,
"eval_validation_agenda_digitale_runtime": 201.6531,
"eval_validation_agenda_digitale_samples_per_second": 1.26,
"eval_validation_agenda_digitale_steps_per_second": 0.63,
"step": 624
},
{
"epoch": 0.21,
"eval_validation_leggepertutti_loss": 1.2475595474243164,
"eval_validation_leggepertutti_runtime": 37.3103,
"eval_validation_leggepertutti_samples_per_second": 1.26,
"eval_validation_leggepertutti_steps_per_second": 0.643,
"step": 624
},
{
"epoch": 0.21,
"eval_validation_altalexprivacy_loss": 1.227779507637024,
"eval_validation_altalexprivacy_runtime": 62.7042,
"eval_validation_altalexprivacy_samples_per_second": 1.26,
"eval_validation_altalexprivacy_steps_per_second": 0.638,
"step": 624
}
],
"logging_steps": 78,
"max_steps": 5888,
"num_input_tokens_seen": 0,
"num_train_epochs": 2,
"save_steps": 78,
"total_flos": 6.978879199490605e+18,
"train_batch_size": 1,
"trial_name": null,
"trial_params": null
}