LexLLMv0.0.0.x.10.15e / trainer_state.json
andrealexroom's picture
Upload folder using huggingface_hub
1ae5865 verified
{
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 0.6464586846543001,
"eval_steps": 187,
"global_step": 7667,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.0,
"grad_norm": 8.5,
"learning_rate": 1e-05,
"loss": 1.9119,
"step": 1
},
{
"epoch": 0.02,
"grad_norm": 4.59375,
"learning_rate": 1e-05,
"loss": 1.7164,
"step": 187
},
{
"epoch": 0.02,
"eval_validation_privacy_sources_loss": 1.0334585905075073,
"eval_validation_privacy_sources_runtime": 212.2756,
"eval_validation_privacy_sources_samples_per_second": 1.804,
"eval_validation_privacy_sources_steps_per_second": 0.904,
"step": 187
},
{
"epoch": 0.02,
"eval_validation_agenda_digitale_loss": 1.5160526037216187,
"eval_validation_agenda_digitale_runtime": 188.4106,
"eval_validation_agenda_digitale_samples_per_second": 1.805,
"eval_validation_agenda_digitale_steps_per_second": 0.902,
"step": 187
},
{
"epoch": 0.02,
"eval_validation_leggepertutti_loss": 1.3417482376098633,
"eval_validation_leggepertutti_runtime": 35.8837,
"eval_validation_leggepertutti_samples_per_second": 1.811,
"eval_validation_leggepertutti_steps_per_second": 0.92,
"step": 187
},
{
"epoch": 0.02,
"eval_validation_altalexprivacy_loss": 1.3672271966934204,
"eval_validation_altalexprivacy_runtime": 58.6063,
"eval_validation_altalexprivacy_samples_per_second": 1.809,
"eval_validation_altalexprivacy_steps_per_second": 0.904,
"step": 187
},
{
"epoch": 0.02,
"eval_validation_mc4_it_loss": 1.9217756986618042,
"eval_validation_mc4_it_runtime": 78.6277,
"eval_validation_mc4_it_samples_per_second": 1.806,
"eval_validation_mc4_it_steps_per_second": 0.903,
"step": 187
},
{
"epoch": 0.02,
"eval_validation_dolma_loss": 1.9627543687820435,
"eval_validation_dolma_runtime": 54.0088,
"eval_validation_dolma_samples_per_second": 1.815,
"eval_validation_dolma_steps_per_second": 0.907,
"step": 187
},
{
"epoch": 0.03,
"grad_norm": 3.859375,
"learning_rate": 1e-05,
"loss": 1.6579,
"step": 374
},
{
"epoch": 0.03,
"eval_validation_privacy_sources_loss": 0.9830030798912048,
"eval_validation_privacy_sources_runtime": 210.811,
"eval_validation_privacy_sources_samples_per_second": 1.817,
"eval_validation_privacy_sources_steps_per_second": 0.911,
"step": 374
},
{
"epoch": 0.03,
"eval_validation_agenda_digitale_loss": 1.4846152067184448,
"eval_validation_agenda_digitale_runtime": 187.1837,
"eval_validation_agenda_digitale_samples_per_second": 1.816,
"eval_validation_agenda_digitale_steps_per_second": 0.908,
"step": 374
},
{
"epoch": 0.03,
"eval_validation_leggepertutti_loss": 1.3172152042388916,
"eval_validation_leggepertutti_runtime": 35.7866,
"eval_validation_leggepertutti_samples_per_second": 1.816,
"eval_validation_leggepertutti_steps_per_second": 0.922,
"step": 374
},
{
"epoch": 0.03,
"eval_validation_altalexprivacy_loss": 1.3326139450073242,
"eval_validation_altalexprivacy_runtime": 58.3381,
"eval_validation_altalexprivacy_samples_per_second": 1.817,
"eval_validation_altalexprivacy_steps_per_second": 0.908,
"step": 374
},
{
"epoch": 0.03,
"eval_validation_mc4_it_loss": 1.8915678262710571,
"eval_validation_mc4_it_runtime": 78.169,
"eval_validation_mc4_it_samples_per_second": 1.817,
"eval_validation_mc4_it_steps_per_second": 0.908,
"step": 374
},
{
"epoch": 0.03,
"eval_validation_dolma_loss": 1.9608409404754639,
"eval_validation_dolma_runtime": 53.9716,
"eval_validation_dolma_samples_per_second": 1.816,
"eval_validation_dolma_steps_per_second": 0.908,
"step": 374
},
{
"epoch": 0.05,
"grad_norm": 3.484375,
"learning_rate": 1e-05,
"loss": 1.6402,
"step": 561
},
{
"epoch": 0.05,
"eval_validation_privacy_sources_loss": 0.9559596180915833,
"eval_validation_privacy_sources_runtime": 210.7246,
"eval_validation_privacy_sources_samples_per_second": 1.818,
"eval_validation_privacy_sources_steps_per_second": 0.911,
"step": 561
},
{
"epoch": 0.05,
"eval_validation_agenda_digitale_loss": 1.4720743894577026,
"eval_validation_agenda_digitale_runtime": 187.11,
"eval_validation_agenda_digitale_samples_per_second": 1.817,
"eval_validation_agenda_digitale_steps_per_second": 0.909,
"step": 561
},
{
"epoch": 0.05,
"eval_validation_leggepertutti_loss": 1.3074837923049927,
"eval_validation_leggepertutti_runtime": 35.7797,
"eval_validation_leggepertutti_samples_per_second": 1.817,
"eval_validation_leggepertutti_steps_per_second": 0.922,
"step": 561
},
{
"epoch": 0.05,
"eval_validation_altalexprivacy_loss": 1.3142317533493042,
"eval_validation_altalexprivacy_runtime": 58.315,
"eval_validation_altalexprivacy_samples_per_second": 1.818,
"eval_validation_altalexprivacy_steps_per_second": 0.909,
"step": 561
},
{
"epoch": 0.05,
"eval_validation_mc4_it_loss": 1.87451171875,
"eval_validation_mc4_it_runtime": 78.1299,
"eval_validation_mc4_it_samples_per_second": 1.817,
"eval_validation_mc4_it_steps_per_second": 0.909,
"step": 561
},
{
"epoch": 0.05,
"eval_validation_dolma_loss": 1.956404447555542,
"eval_validation_dolma_runtime": 53.9194,
"eval_validation_dolma_samples_per_second": 1.818,
"eval_validation_dolma_steps_per_second": 0.909,
"step": 561
},
{
"epoch": 0.06,
"grad_norm": 9.625,
"learning_rate": 1e-05,
"loss": 1.6284,
"step": 748
},
{
"epoch": 0.06,
"eval_validation_privacy_sources_loss": 0.933682918548584,
"eval_validation_privacy_sources_runtime": 210.799,
"eval_validation_privacy_sources_samples_per_second": 1.817,
"eval_validation_privacy_sources_steps_per_second": 0.911,
"step": 748
},
{
"epoch": 0.06,
"eval_validation_agenda_digitale_loss": 1.4578524827957153,
"eval_validation_agenda_digitale_runtime": 187.08,
"eval_validation_agenda_digitale_samples_per_second": 1.817,
"eval_validation_agenda_digitale_steps_per_second": 0.909,
"step": 748
},
{
"epoch": 0.06,
"eval_validation_leggepertutti_loss": 1.2905200719833374,
"eval_validation_leggepertutti_runtime": 35.7738,
"eval_validation_leggepertutti_samples_per_second": 1.817,
"eval_validation_leggepertutti_steps_per_second": 0.922,
"step": 748
},
{
"epoch": 0.06,
"eval_validation_altalexprivacy_loss": 1.2977306842803955,
"eval_validation_altalexprivacy_runtime": 58.3215,
"eval_validation_altalexprivacy_samples_per_second": 1.818,
"eval_validation_altalexprivacy_steps_per_second": 0.909,
"step": 748
},
{
"epoch": 0.06,
"eval_validation_mc4_it_loss": 1.8592724800109863,
"eval_validation_mc4_it_runtime": 78.1141,
"eval_validation_mc4_it_samples_per_second": 1.818,
"eval_validation_mc4_it_steps_per_second": 0.909,
"step": 748
},
{
"epoch": 0.06,
"eval_validation_dolma_loss": 1.9574466943740845,
"eval_validation_dolma_runtime": 53.9297,
"eval_validation_dolma_samples_per_second": 1.817,
"eval_validation_dolma_steps_per_second": 0.909,
"step": 748
},
{
"epoch": 0.08,
"grad_norm": 3.390625,
"learning_rate": 1e-05,
"loss": 1.6082,
"step": 935
},
{
"epoch": 0.08,
"eval_validation_privacy_sources_loss": 0.919241726398468,
"eval_validation_privacy_sources_runtime": 210.6302,
"eval_validation_privacy_sources_samples_per_second": 1.818,
"eval_validation_privacy_sources_steps_per_second": 0.912,
"step": 935
},
{
"epoch": 0.08,
"eval_validation_agenda_digitale_loss": 1.4438046216964722,
"eval_validation_agenda_digitale_runtime": 187.0727,
"eval_validation_agenda_digitale_samples_per_second": 1.817,
"eval_validation_agenda_digitale_steps_per_second": 0.909,
"step": 935
},
{
"epoch": 0.08,
"eval_validation_leggepertutti_loss": 1.2822848558425903,
"eval_validation_leggepertutti_runtime": 35.7909,
"eval_validation_leggepertutti_samples_per_second": 1.816,
"eval_validation_leggepertutti_steps_per_second": 0.922,
"step": 935
},
{
"epoch": 0.08,
"eval_validation_altalexprivacy_loss": 1.2801928520202637,
"eval_validation_altalexprivacy_runtime": 58.3369,
"eval_validation_altalexprivacy_samples_per_second": 1.817,
"eval_validation_altalexprivacy_steps_per_second": 0.909,
"step": 935
},
{
"epoch": 0.08,
"eval_validation_mc4_it_loss": 1.8457767963409424,
"eval_validation_mc4_it_runtime": 78.1714,
"eval_validation_mc4_it_samples_per_second": 1.817,
"eval_validation_mc4_it_steps_per_second": 0.908,
"step": 935
},
{
"epoch": 0.08,
"eval_validation_dolma_loss": 1.9588344097137451,
"eval_validation_dolma_runtime": 53.9317,
"eval_validation_dolma_samples_per_second": 1.817,
"eval_validation_dolma_steps_per_second": 0.909,
"step": 935
},
{
"epoch": 0.09,
"grad_norm": 3.46875,
"learning_rate": 1e-05,
"loss": 1.5989,
"step": 1122
},
{
"epoch": 0.09,
"eval_validation_privacy_sources_loss": 0.9032600522041321,
"eval_validation_privacy_sources_runtime": 210.7002,
"eval_validation_privacy_sources_samples_per_second": 1.818,
"eval_validation_privacy_sources_steps_per_second": 0.911,
"step": 1122
},
{
"epoch": 0.09,
"eval_validation_agenda_digitale_loss": 1.433292031288147,
"eval_validation_agenda_digitale_runtime": 187.0521,
"eval_validation_agenda_digitale_samples_per_second": 1.818,
"eval_validation_agenda_digitale_steps_per_second": 0.909,
"step": 1122
},
{
"epoch": 0.09,
"eval_validation_leggepertutti_loss": 1.2763991355895996,
"eval_validation_leggepertutti_runtime": 35.7627,
"eval_validation_leggepertutti_samples_per_second": 1.818,
"eval_validation_leggepertutti_steps_per_second": 0.923,
"step": 1122
},
{
"epoch": 0.09,
"eval_validation_altalexprivacy_loss": 1.2663257122039795,
"eval_validation_altalexprivacy_runtime": 58.3197,
"eval_validation_altalexprivacy_samples_per_second": 1.818,
"eval_validation_altalexprivacy_steps_per_second": 0.909,
"step": 1122
},
{
"epoch": 0.09,
"eval_validation_mc4_it_loss": 1.8358547687530518,
"eval_validation_mc4_it_runtime": 78.1183,
"eval_validation_mc4_it_samples_per_second": 1.818,
"eval_validation_mc4_it_steps_per_second": 0.909,
"step": 1122
},
{
"epoch": 0.09,
"eval_validation_dolma_loss": 2.0097882747650146,
"eval_validation_dolma_runtime": 53.9266,
"eval_validation_dolma_samples_per_second": 1.817,
"eval_validation_dolma_steps_per_second": 0.909,
"step": 1122
},
{
"epoch": 0.11,
"grad_norm": 3.40625,
"learning_rate": 1e-05,
"loss": 1.6161,
"step": 1309
},
{
"epoch": 0.11,
"eval_validation_privacy_sources_loss": 0.8906105160713196,
"eval_validation_privacy_sources_runtime": 210.7431,
"eval_validation_privacy_sources_samples_per_second": 1.817,
"eval_validation_privacy_sources_steps_per_second": 0.911,
"step": 1309
},
{
"epoch": 0.11,
"eval_validation_agenda_digitale_loss": 1.4252179861068726,
"eval_validation_agenda_digitale_runtime": 187.1263,
"eval_validation_agenda_digitale_samples_per_second": 1.817,
"eval_validation_agenda_digitale_steps_per_second": 0.908,
"step": 1309
},
{
"epoch": 0.11,
"eval_validation_leggepertutti_loss": 1.2690521478652954,
"eval_validation_leggepertutti_runtime": 35.7673,
"eval_validation_leggepertutti_samples_per_second": 1.817,
"eval_validation_leggepertutti_steps_per_second": 0.923,
"step": 1309
},
{
"epoch": 0.11,
"eval_validation_altalexprivacy_loss": 1.258312463760376,
"eval_validation_altalexprivacy_runtime": 58.3303,
"eval_validation_altalexprivacy_samples_per_second": 1.817,
"eval_validation_altalexprivacy_steps_per_second": 0.909,
"step": 1309
},
{
"epoch": 0.11,
"eval_validation_mc4_it_loss": 1.8263970613479614,
"eval_validation_mc4_it_runtime": 78.147,
"eval_validation_mc4_it_samples_per_second": 1.817,
"eval_validation_mc4_it_steps_per_second": 0.909,
"step": 1309
},
{
"epoch": 0.11,
"eval_validation_dolma_loss": 1.9555673599243164,
"eval_validation_dolma_runtime": 53.9288,
"eval_validation_dolma_samples_per_second": 1.817,
"eval_validation_dolma_steps_per_second": 0.909,
"step": 1309
},
{
"epoch": 0.13,
"grad_norm": 3.25,
"learning_rate": 1e-05,
"loss": 1.5911,
"step": 1496
},
{
"epoch": 0.13,
"eval_validation_privacy_sources_loss": 0.8815954327583313,
"eval_validation_privacy_sources_runtime": 210.7215,
"eval_validation_privacy_sources_samples_per_second": 1.818,
"eval_validation_privacy_sources_steps_per_second": 0.911,
"step": 1496
},
{
"epoch": 0.13,
"eval_validation_agenda_digitale_loss": 1.4188568592071533,
"eval_validation_agenda_digitale_runtime": 187.0962,
"eval_validation_agenda_digitale_samples_per_second": 1.817,
"eval_validation_agenda_digitale_steps_per_second": 0.909,
"step": 1496
},
{
"epoch": 0.13,
"eval_validation_leggepertutti_loss": 1.269508957862854,
"eval_validation_leggepertutti_runtime": 35.7762,
"eval_validation_leggepertutti_samples_per_second": 1.817,
"eval_validation_leggepertutti_steps_per_second": 0.922,
"step": 1496
},
{
"epoch": 0.13,
"eval_validation_altalexprivacy_loss": 1.2556049823760986,
"eval_validation_altalexprivacy_runtime": 58.3467,
"eval_validation_altalexprivacy_samples_per_second": 1.817,
"eval_validation_altalexprivacy_steps_per_second": 0.908,
"step": 1496
},
{
"epoch": 0.13,
"eval_validation_mc4_it_loss": 1.8174842596054077,
"eval_validation_mc4_it_runtime": 78.1412,
"eval_validation_mc4_it_samples_per_second": 1.817,
"eval_validation_mc4_it_steps_per_second": 0.909,
"step": 1496
},
{
"epoch": 0.13,
"eval_validation_dolma_loss": 1.9592047929763794,
"eval_validation_dolma_runtime": 53.9471,
"eval_validation_dolma_samples_per_second": 1.817,
"eval_validation_dolma_steps_per_second": 0.908,
"step": 1496
},
{
"epoch": 0.14,
"grad_norm": 3.359375,
"learning_rate": 1e-05,
"loss": 1.5694,
"step": 1683
},
{
"epoch": 0.14,
"eval_validation_privacy_sources_loss": 0.8737432956695557,
"eval_validation_privacy_sources_runtime": 210.8196,
"eval_validation_privacy_sources_samples_per_second": 1.817,
"eval_validation_privacy_sources_steps_per_second": 0.911,
"step": 1683
},
{
"epoch": 0.14,
"eval_validation_agenda_digitale_loss": 1.4132717847824097,
"eval_validation_agenda_digitale_runtime": 187.1554,
"eval_validation_agenda_digitale_samples_per_second": 1.817,
"eval_validation_agenda_digitale_steps_per_second": 0.908,
"step": 1683
},
{
"epoch": 0.14,
"eval_validation_leggepertutti_loss": 1.2630620002746582,
"eval_validation_leggepertutti_runtime": 35.7934,
"eval_validation_leggepertutti_samples_per_second": 1.816,
"eval_validation_leggepertutti_steps_per_second": 0.922,
"step": 1683
},
{
"epoch": 0.14,
"eval_validation_altalexprivacy_loss": 1.247528076171875,
"eval_validation_altalexprivacy_runtime": 58.3547,
"eval_validation_altalexprivacy_samples_per_second": 1.816,
"eval_validation_altalexprivacy_steps_per_second": 0.908,
"step": 1683
},
{
"epoch": 0.14,
"eval_validation_mc4_it_loss": 1.8103784322738647,
"eval_validation_mc4_it_runtime": 78.1665,
"eval_validation_mc4_it_samples_per_second": 1.817,
"eval_validation_mc4_it_steps_per_second": 0.908,
"step": 1683
},
{
"epoch": 0.14,
"eval_validation_dolma_loss": 1.956905722618103,
"eval_validation_dolma_runtime": 53.9622,
"eval_validation_dolma_samples_per_second": 1.816,
"eval_validation_dolma_steps_per_second": 0.908,
"step": 1683
},
{
"epoch": 0.16,
"grad_norm": 3.1875,
"learning_rate": 1e-05,
"loss": 1.563,
"step": 1870
},
{
"epoch": 0.16,
"eval_validation_privacy_sources_loss": 0.8650386929512024,
"eval_validation_privacy_sources_runtime": 210.8215,
"eval_validation_privacy_sources_samples_per_second": 1.817,
"eval_validation_privacy_sources_steps_per_second": 0.911,
"step": 1870
},
{
"epoch": 0.16,
"eval_validation_agenda_digitale_loss": 1.4071804285049438,
"eval_validation_agenda_digitale_runtime": 187.1502,
"eval_validation_agenda_digitale_samples_per_second": 1.817,
"eval_validation_agenda_digitale_steps_per_second": 0.908,
"step": 1870
},
{
"epoch": 0.16,
"eval_validation_leggepertutti_loss": 1.2562586069107056,
"eval_validation_leggepertutti_runtime": 35.7742,
"eval_validation_leggepertutti_samples_per_second": 1.817,
"eval_validation_leggepertutti_steps_per_second": 0.922,
"step": 1870
},
{
"epoch": 0.16,
"eval_validation_altalexprivacy_loss": 1.240662932395935,
"eval_validation_altalexprivacy_runtime": 58.3499,
"eval_validation_altalexprivacy_samples_per_second": 1.817,
"eval_validation_altalexprivacy_steps_per_second": 0.908,
"step": 1870
},
{
"epoch": 0.16,
"eval_validation_mc4_it_loss": 1.8042465448379517,
"eval_validation_mc4_it_runtime": 78.1652,
"eval_validation_mc4_it_samples_per_second": 1.817,
"eval_validation_mc4_it_steps_per_second": 0.908,
"step": 1870
},
{
"epoch": 0.16,
"eval_validation_dolma_loss": 1.9527182579040527,
"eval_validation_dolma_runtime": 53.948,
"eval_validation_dolma_samples_per_second": 1.817,
"eval_validation_dolma_steps_per_second": 0.908,
"step": 1870
},
{
"epoch": 0.17,
"grad_norm": 3.171875,
"learning_rate": 1e-05,
"loss": 1.5747,
"step": 2057
},
{
"epoch": 0.17,
"eval_validation_privacy_sources_loss": 0.8606281876564026,
"eval_validation_privacy_sources_runtime": 210.8197,
"eval_validation_privacy_sources_samples_per_second": 1.817,
"eval_validation_privacy_sources_steps_per_second": 0.911,
"step": 2057
},
{
"epoch": 0.17,
"eval_validation_agenda_digitale_loss": 1.401216983795166,
"eval_validation_agenda_digitale_runtime": 187.1929,
"eval_validation_agenda_digitale_samples_per_second": 1.816,
"eval_validation_agenda_digitale_steps_per_second": 0.908,
"step": 2057
},
{
"epoch": 0.17,
"eval_validation_leggepertutti_loss": 1.249880075454712,
"eval_validation_leggepertutti_runtime": 35.7828,
"eval_validation_leggepertutti_samples_per_second": 1.817,
"eval_validation_leggepertutti_steps_per_second": 0.922,
"step": 2057
},
{
"epoch": 0.17,
"eval_validation_altalexprivacy_loss": 1.2347912788391113,
"eval_validation_altalexprivacy_runtime": 58.3271,
"eval_validation_altalexprivacy_samples_per_second": 1.817,
"eval_validation_altalexprivacy_steps_per_second": 0.909,
"step": 2057
},
{
"epoch": 0.17,
"eval_validation_mc4_it_loss": 1.7988897562026978,
"eval_validation_mc4_it_runtime": 78.1637,
"eval_validation_mc4_it_samples_per_second": 1.817,
"eval_validation_mc4_it_steps_per_second": 0.908,
"step": 2057
},
{
"epoch": 0.17,
"eval_validation_dolma_loss": 1.9512945413589478,
"eval_validation_dolma_runtime": 53.9545,
"eval_validation_dolma_samples_per_second": 1.816,
"eval_validation_dolma_steps_per_second": 0.908,
"step": 2057
},
{
"epoch": 0.19,
"grad_norm": 3.703125,
"learning_rate": 1e-05,
"loss": 1.5631,
"step": 2244
},
{
"epoch": 0.19,
"eval_validation_privacy_sources_loss": 0.8552854061126709,
"eval_validation_privacy_sources_runtime": 210.7654,
"eval_validation_privacy_sources_samples_per_second": 1.817,
"eval_validation_privacy_sources_steps_per_second": 0.911,
"step": 2244
},
{
"epoch": 0.19,
"eval_validation_agenda_digitale_loss": 1.3981597423553467,
"eval_validation_agenda_digitale_runtime": 187.0896,
"eval_validation_agenda_digitale_samples_per_second": 1.817,
"eval_validation_agenda_digitale_steps_per_second": 0.909,
"step": 2244
},
{
"epoch": 0.19,
"eval_validation_leggepertutti_loss": 1.249776840209961,
"eval_validation_leggepertutti_runtime": 35.7701,
"eval_validation_leggepertutti_samples_per_second": 1.817,
"eval_validation_leggepertutti_steps_per_second": 0.923,
"step": 2244
},
{
"epoch": 0.19,
"eval_validation_altalexprivacy_loss": 1.2317657470703125,
"eval_validation_altalexprivacy_runtime": 58.3402,
"eval_validation_altalexprivacy_samples_per_second": 1.817,
"eval_validation_altalexprivacy_steps_per_second": 0.908,
"step": 2244
},
{
"epoch": 0.19,
"eval_validation_mc4_it_loss": 1.7936795949935913,
"eval_validation_mc4_it_runtime": 78.1539,
"eval_validation_mc4_it_samples_per_second": 1.817,
"eval_validation_mc4_it_steps_per_second": 0.908,
"step": 2244
},
{
"epoch": 0.19,
"eval_validation_dolma_loss": 1.949507474899292,
"eval_validation_dolma_runtime": 53.9287,
"eval_validation_dolma_samples_per_second": 1.817,
"eval_validation_dolma_steps_per_second": 0.909,
"step": 2244
},
{
"epoch": 0.2,
"grad_norm": 3.28125,
"learning_rate": 1e-05,
"loss": 1.5502,
"step": 2431
},
{
"epoch": 0.2,
"eval_validation_privacy_sources_loss": 0.8490111231803894,
"eval_validation_privacy_sources_runtime": 210.7371,
"eval_validation_privacy_sources_samples_per_second": 1.817,
"eval_validation_privacy_sources_steps_per_second": 0.911,
"step": 2431
},
{
"epoch": 0.2,
"eval_validation_agenda_digitale_loss": 1.391976237297058,
"eval_validation_agenda_digitale_runtime": 187.0694,
"eval_validation_agenda_digitale_samples_per_second": 1.818,
"eval_validation_agenda_digitale_steps_per_second": 0.909,
"step": 2431
},
{
"epoch": 0.2,
"eval_validation_leggepertutti_loss": 1.2457927465438843,
"eval_validation_leggepertutti_runtime": 35.7756,
"eval_validation_leggepertutti_samples_per_second": 1.817,
"eval_validation_leggepertutti_steps_per_second": 0.922,
"step": 2431
},
{
"epoch": 0.2,
"eval_validation_altalexprivacy_loss": 1.2249823808670044,
"eval_validation_altalexprivacy_runtime": 58.3391,
"eval_validation_altalexprivacy_samples_per_second": 1.817,
"eval_validation_altalexprivacy_steps_per_second": 0.908,
"step": 2431
},
{
"epoch": 0.2,
"eval_validation_mc4_it_loss": 1.7877947092056274,
"eval_validation_mc4_it_runtime": 78.1519,
"eval_validation_mc4_it_samples_per_second": 1.817,
"eval_validation_mc4_it_steps_per_second": 0.908,
"step": 2431
},
{
"epoch": 0.2,
"eval_validation_dolma_loss": 1.9478541612625122,
"eval_validation_dolma_runtime": 53.9392,
"eval_validation_dolma_samples_per_second": 1.817,
"eval_validation_dolma_steps_per_second": 0.908,
"step": 2431
},
{
"epoch": 0.22,
"grad_norm": 3.109375,
"learning_rate": 1e-05,
"loss": 1.5612,
"step": 2618
},
{
"epoch": 0.22,
"eval_validation_privacy_sources_loss": 0.8443201184272766,
"eval_validation_privacy_sources_runtime": 210.7053,
"eval_validation_privacy_sources_samples_per_second": 1.818,
"eval_validation_privacy_sources_steps_per_second": 0.911,
"step": 2618
},
{
"epoch": 0.22,
"eval_validation_agenda_digitale_loss": 1.3874356746673584,
"eval_validation_agenda_digitale_runtime": 187.0741,
"eval_validation_agenda_digitale_samples_per_second": 1.817,
"eval_validation_agenda_digitale_steps_per_second": 0.909,
"step": 2618
},
{
"epoch": 0.22,
"eval_validation_leggepertutti_loss": 1.242628574371338,
"eval_validation_leggepertutti_runtime": 35.7724,
"eval_validation_leggepertutti_samples_per_second": 1.817,
"eval_validation_leggepertutti_steps_per_second": 0.922,
"step": 2618
},
{
"epoch": 0.22,
"eval_validation_altalexprivacy_loss": 1.22023606300354,
"eval_validation_altalexprivacy_runtime": 58.3215,
"eval_validation_altalexprivacy_samples_per_second": 1.818,
"eval_validation_altalexprivacy_steps_per_second": 0.909,
"step": 2618
},
{
"epoch": 0.22,
"eval_validation_mc4_it_loss": 1.783267617225647,
"eval_validation_mc4_it_runtime": 78.1277,
"eval_validation_mc4_it_samples_per_second": 1.818,
"eval_validation_mc4_it_steps_per_second": 0.909,
"step": 2618
},
{
"epoch": 0.22,
"eval_validation_dolma_loss": 1.9474639892578125,
"eval_validation_dolma_runtime": 53.9215,
"eval_validation_dolma_samples_per_second": 1.817,
"eval_validation_dolma_steps_per_second": 0.909,
"step": 2618
},
{
"epoch": 0.24,
"grad_norm": 9.0,
"learning_rate": 1e-05,
"loss": 1.5625,
"step": 2805
},
{
"epoch": 0.24,
"eval_validation_privacy_sources_loss": 0.8386737108230591,
"eval_validation_privacy_sources_runtime": 210.6496,
"eval_validation_privacy_sources_samples_per_second": 1.818,
"eval_validation_privacy_sources_steps_per_second": 0.911,
"step": 2805
},
{
"epoch": 0.24,
"eval_validation_agenda_digitale_loss": 1.3825620412826538,
"eval_validation_agenda_digitale_runtime": 186.9852,
"eval_validation_agenda_digitale_samples_per_second": 1.818,
"eval_validation_agenda_digitale_steps_per_second": 0.909,
"step": 2805
},
{
"epoch": 0.24,
"eval_validation_leggepertutti_loss": 1.2439346313476562,
"eval_validation_leggepertutti_runtime": 35.762,
"eval_validation_leggepertutti_samples_per_second": 1.818,
"eval_validation_leggepertutti_steps_per_second": 0.923,
"step": 2805
},
{
"epoch": 0.24,
"eval_validation_altalexprivacy_loss": 1.2144984006881714,
"eval_validation_altalexprivacy_runtime": 58.2929,
"eval_validation_altalexprivacy_samples_per_second": 1.818,
"eval_validation_altalexprivacy_steps_per_second": 0.909,
"step": 2805
},
{
"epoch": 0.24,
"eval_validation_mc4_it_loss": 1.7777953147888184,
"eval_validation_mc4_it_runtime": 78.0862,
"eval_validation_mc4_it_samples_per_second": 1.819,
"eval_validation_mc4_it_steps_per_second": 0.909,
"step": 2805
},
{
"epoch": 0.24,
"eval_validation_dolma_loss": 1.9426158666610718,
"eval_validation_dolma_runtime": 53.8932,
"eval_validation_dolma_samples_per_second": 1.818,
"eval_validation_dolma_steps_per_second": 0.909,
"step": 2805
},
{
"epoch": 0.25,
"grad_norm": 3.09375,
"learning_rate": 1e-05,
"loss": 1.5659,
"step": 2992
},
{
"epoch": 0.25,
"eval_validation_privacy_sources_loss": 0.8341156840324402,
"eval_validation_privacy_sources_runtime": 210.6733,
"eval_validation_privacy_sources_samples_per_second": 1.818,
"eval_validation_privacy_sources_steps_per_second": 0.911,
"step": 2992
},
{
"epoch": 0.25,
"eval_validation_agenda_digitale_loss": 1.3803825378417969,
"eval_validation_agenda_digitale_runtime": 187.0096,
"eval_validation_agenda_digitale_samples_per_second": 1.818,
"eval_validation_agenda_digitale_steps_per_second": 0.909,
"step": 2992
},
{
"epoch": 0.25,
"eval_validation_leggepertutti_loss": 1.2398289442062378,
"eval_validation_leggepertutti_runtime": 35.7578,
"eval_validation_leggepertutti_samples_per_second": 1.818,
"eval_validation_leggepertutti_steps_per_second": 0.923,
"step": 2992
},
{
"epoch": 0.25,
"eval_validation_altalexprivacy_loss": 1.2100013494491577,
"eval_validation_altalexprivacy_runtime": 58.3051,
"eval_validation_altalexprivacy_samples_per_second": 1.818,
"eval_validation_altalexprivacy_steps_per_second": 0.909,
"step": 2992
},
{
"epoch": 0.25,
"eval_validation_mc4_it_loss": 1.7749333381652832,
"eval_validation_mc4_it_runtime": 78.0968,
"eval_validation_mc4_it_samples_per_second": 1.818,
"eval_validation_mc4_it_steps_per_second": 0.909,
"step": 2992
},
{
"epoch": 0.25,
"eval_validation_dolma_loss": 1.9426519870758057,
"eval_validation_dolma_runtime": 53.9195,
"eval_validation_dolma_samples_per_second": 1.818,
"eval_validation_dolma_steps_per_second": 0.909,
"step": 2992
},
{
"epoch": 0.27,
"grad_norm": 3.265625,
"learning_rate": 1e-05,
"loss": 1.5527,
"step": 3179
},
{
"epoch": 0.27,
"eval_validation_privacy_sources_loss": 0.8318145871162415,
"eval_validation_privacy_sources_runtime": 210.7255,
"eval_validation_privacy_sources_samples_per_second": 1.818,
"eval_validation_privacy_sources_steps_per_second": 0.911,
"step": 3179
},
{
"epoch": 0.27,
"eval_validation_agenda_digitale_loss": 1.378063440322876,
"eval_validation_agenda_digitale_runtime": 187.087,
"eval_validation_agenda_digitale_samples_per_second": 1.817,
"eval_validation_agenda_digitale_steps_per_second": 0.909,
"step": 3179
},
{
"epoch": 0.27,
"eval_validation_leggepertutti_loss": 1.2370458841323853,
"eval_validation_leggepertutti_runtime": 35.7809,
"eval_validation_leggepertutti_samples_per_second": 1.817,
"eval_validation_leggepertutti_steps_per_second": 0.922,
"step": 3179
},
{
"epoch": 0.27,
"eval_validation_altalexprivacy_loss": 1.2080256938934326,
"eval_validation_altalexprivacy_runtime": 58.3354,
"eval_validation_altalexprivacy_samples_per_second": 1.817,
"eval_validation_altalexprivacy_steps_per_second": 0.909,
"step": 3179
},
{
"epoch": 0.27,
"eval_validation_mc4_it_loss": 1.7702537775039673,
"eval_validation_mc4_it_runtime": 78.1409,
"eval_validation_mc4_it_samples_per_second": 1.817,
"eval_validation_mc4_it_steps_per_second": 0.909,
"step": 3179
},
{
"epoch": 0.27,
"eval_validation_dolma_loss": 1.950868844985962,
"eval_validation_dolma_runtime": 53.9369,
"eval_validation_dolma_samples_per_second": 1.817,
"eval_validation_dolma_steps_per_second": 0.908,
"step": 3179
},
{
"epoch": 0.28,
"grad_norm": 3.265625,
"learning_rate": 1e-05,
"loss": 1.5476,
"step": 3366
},
{
"epoch": 0.28,
"eval_validation_privacy_sources_loss": 0.8264653086662292,
"eval_validation_privacy_sources_runtime": 210.7205,
"eval_validation_privacy_sources_samples_per_second": 1.818,
"eval_validation_privacy_sources_steps_per_second": 0.911,
"step": 3366
},
{
"epoch": 0.28,
"eval_validation_agenda_digitale_loss": 1.3743617534637451,
"eval_validation_agenda_digitale_runtime": 187.1448,
"eval_validation_agenda_digitale_samples_per_second": 1.817,
"eval_validation_agenda_digitale_steps_per_second": 0.908,
"step": 3366
},
{
"epoch": 0.28,
"eval_validation_leggepertutti_loss": 1.2376344203948975,
"eval_validation_leggepertutti_runtime": 35.7729,
"eval_validation_leggepertutti_samples_per_second": 1.817,
"eval_validation_leggepertutti_steps_per_second": 0.922,
"step": 3366
},
{
"epoch": 0.28,
"eval_validation_altalexprivacy_loss": 1.2052847146987915,
"eval_validation_altalexprivacy_runtime": 58.3348,
"eval_validation_altalexprivacy_samples_per_second": 1.817,
"eval_validation_altalexprivacy_steps_per_second": 0.909,
"step": 3366
},
{
"epoch": 0.28,
"eval_validation_mc4_it_loss": 1.7669274806976318,
"eval_validation_mc4_it_runtime": 78.1532,
"eval_validation_mc4_it_samples_per_second": 1.817,
"eval_validation_mc4_it_steps_per_second": 0.908,
"step": 3366
},
{
"epoch": 0.28,
"eval_validation_dolma_loss": 1.9461361169815063,
"eval_validation_dolma_runtime": 53.9281,
"eval_validation_dolma_samples_per_second": 1.817,
"eval_validation_dolma_steps_per_second": 0.909,
"step": 3366
},
{
"epoch": 0.3,
"grad_norm": 3.625,
"learning_rate": 1e-05,
"loss": 1.558,
"step": 3553
},
{
"epoch": 0.3,
"eval_validation_privacy_sources_loss": 0.822688102722168,
"eval_validation_privacy_sources_runtime": 210.9998,
"eval_validation_privacy_sources_samples_per_second": 1.815,
"eval_validation_privacy_sources_steps_per_second": 0.91,
"step": 3553
},
{
"epoch": 0.3,
"eval_validation_agenda_digitale_loss": 1.3728498220443726,
"eval_validation_agenda_digitale_runtime": 187.2711,
"eval_validation_agenda_digitale_samples_per_second": 1.816,
"eval_validation_agenda_digitale_steps_per_second": 0.908,
"step": 3553
},
{
"epoch": 0.3,
"eval_validation_leggepertutti_loss": 1.2360832691192627,
"eval_validation_leggepertutti_runtime": 35.7958,
"eval_validation_leggepertutti_samples_per_second": 1.816,
"eval_validation_leggepertutti_steps_per_second": 0.922,
"step": 3553
},
{
"epoch": 0.3,
"eval_validation_altalexprivacy_loss": 1.2027132511138916,
"eval_validation_altalexprivacy_runtime": 58.3972,
"eval_validation_altalexprivacy_samples_per_second": 1.815,
"eval_validation_altalexprivacy_steps_per_second": 0.908,
"step": 3553
},
{
"epoch": 0.3,
"eval_validation_mc4_it_loss": 1.7643710374832153,
"eval_validation_mc4_it_runtime": 78.2118,
"eval_validation_mc4_it_samples_per_second": 1.816,
"eval_validation_mc4_it_steps_per_second": 0.908,
"step": 3553
},
{
"epoch": 0.3,
"eval_validation_dolma_loss": 1.9442397356033325,
"eval_validation_dolma_runtime": 53.9609,
"eval_validation_dolma_samples_per_second": 1.816,
"eval_validation_dolma_steps_per_second": 0.908,
"step": 3553
},
{
"epoch": 0.32,
"grad_norm": 4.65625,
"learning_rate": 1e-05,
"loss": 1.535,
"step": 3740
},
{
"epoch": 0.32,
"eval_validation_privacy_sources_loss": 0.8193829655647278,
"eval_validation_privacy_sources_runtime": 210.8214,
"eval_validation_privacy_sources_samples_per_second": 1.817,
"eval_validation_privacy_sources_steps_per_second": 0.911,
"step": 3740
},
{
"epoch": 0.32,
"eval_validation_agenda_digitale_loss": 1.3690398931503296,
"eval_validation_agenda_digitale_runtime": 187.1433,
"eval_validation_agenda_digitale_samples_per_second": 1.817,
"eval_validation_agenda_digitale_steps_per_second": 0.908,
"step": 3740
},
{
"epoch": 0.32,
"eval_validation_leggepertutti_loss": 1.2321481704711914,
"eval_validation_leggepertutti_runtime": 35.7856,
"eval_validation_leggepertutti_samples_per_second": 1.816,
"eval_validation_leggepertutti_steps_per_second": 0.922,
"step": 3740
},
{
"epoch": 0.32,
"eval_validation_altalexprivacy_loss": 1.1964303255081177,
"eval_validation_altalexprivacy_runtime": 58.3552,
"eval_validation_altalexprivacy_samples_per_second": 1.816,
"eval_validation_altalexprivacy_steps_per_second": 0.908,
"step": 3740
},
{
"epoch": 0.32,
"eval_validation_mc4_it_loss": 1.7612253427505493,
"eval_validation_mc4_it_runtime": 78.1715,
"eval_validation_mc4_it_samples_per_second": 1.817,
"eval_validation_mc4_it_steps_per_second": 0.908,
"step": 3740
},
{
"epoch": 0.32,
"eval_validation_dolma_loss": 1.9415688514709473,
"eval_validation_dolma_runtime": 53.9531,
"eval_validation_dolma_samples_per_second": 1.816,
"eval_validation_dolma_steps_per_second": 0.908,
"step": 3740
},
{
"epoch": 0.33,
"grad_norm": 3.203125,
"learning_rate": 1e-05,
"loss": 1.551,
"step": 3927
},
{
"epoch": 0.33,
"eval_validation_privacy_sources_loss": 0.8159863352775574,
"eval_validation_privacy_sources_runtime": 210.8294,
"eval_validation_privacy_sources_samples_per_second": 1.817,
"eval_validation_privacy_sources_steps_per_second": 0.911,
"step": 3927
},
{
"epoch": 0.33,
"eval_validation_agenda_digitale_loss": 1.3660005331039429,
"eval_validation_agenda_digitale_runtime": 187.2084,
"eval_validation_agenda_digitale_samples_per_second": 1.816,
"eval_validation_agenda_digitale_steps_per_second": 0.908,
"step": 3927
},
{
"epoch": 0.33,
"eval_validation_leggepertutti_loss": 1.2291594743728638,
"eval_validation_leggepertutti_runtime": 35.7917,
"eval_validation_leggepertutti_samples_per_second": 1.816,
"eval_validation_leggepertutti_steps_per_second": 0.922,
"step": 3927
},
{
"epoch": 0.33,
"eval_validation_altalexprivacy_loss": 1.1953657865524292,
"eval_validation_altalexprivacy_runtime": 58.3486,
"eval_validation_altalexprivacy_samples_per_second": 1.817,
"eval_validation_altalexprivacy_steps_per_second": 0.908,
"step": 3927
},
{
"epoch": 0.33,
"eval_validation_mc4_it_loss": 1.7568954229354858,
"eval_validation_mc4_it_runtime": 78.1754,
"eval_validation_mc4_it_samples_per_second": 1.816,
"eval_validation_mc4_it_steps_per_second": 0.908,
"step": 3927
},
{
"epoch": 0.33,
"eval_validation_dolma_loss": 1.9388848543167114,
"eval_validation_dolma_runtime": 53.9683,
"eval_validation_dolma_samples_per_second": 1.816,
"eval_validation_dolma_steps_per_second": 0.908,
"step": 3927
},
{
"epoch": 0.35,
"grad_norm": 3.578125,
"learning_rate": 1e-05,
"loss": 1.5452,
"step": 4114
},
{
"epoch": 0.35,
"eval_validation_privacy_sources_loss": 0.8137150406837463,
"eval_validation_privacy_sources_runtime": 210.7357,
"eval_validation_privacy_sources_samples_per_second": 1.817,
"eval_validation_privacy_sources_steps_per_second": 0.911,
"step": 4114
},
{
"epoch": 0.35,
"eval_validation_agenda_digitale_loss": 1.3656136989593506,
"eval_validation_agenda_digitale_runtime": 187.146,
"eval_validation_agenda_digitale_samples_per_second": 1.817,
"eval_validation_agenda_digitale_steps_per_second": 0.908,
"step": 4114
},
{
"epoch": 0.35,
"eval_validation_leggepertutti_loss": 1.2296239137649536,
"eval_validation_leggepertutti_runtime": 35.7766,
"eval_validation_leggepertutti_samples_per_second": 1.817,
"eval_validation_leggepertutti_steps_per_second": 0.922,
"step": 4114
},
{
"epoch": 0.35,
"eval_validation_altalexprivacy_loss": 1.1915565729141235,
"eval_validation_altalexprivacy_runtime": 58.3322,
"eval_validation_altalexprivacy_samples_per_second": 1.817,
"eval_validation_altalexprivacy_steps_per_second": 0.909,
"step": 4114
},
{
"epoch": 0.35,
"eval_validation_mc4_it_loss": 1.7542990446090698,
"eval_validation_mc4_it_runtime": 78.1622,
"eval_validation_mc4_it_samples_per_second": 1.817,
"eval_validation_mc4_it_steps_per_second": 0.908,
"step": 4114
},
{
"epoch": 0.35,
"eval_validation_dolma_loss": 1.9385536909103394,
"eval_validation_dolma_runtime": 53.9474,
"eval_validation_dolma_samples_per_second": 1.817,
"eval_validation_dolma_steps_per_second": 0.908,
"step": 4114
},
{
"epoch": 0.36,
"grad_norm": 3.421875,
"learning_rate": 1e-05,
"loss": 1.5371,
"step": 4301
},
{
"epoch": 0.36,
"eval_validation_privacy_sources_loss": 0.8114846348762512,
"eval_validation_privacy_sources_runtime": 210.7483,
"eval_validation_privacy_sources_samples_per_second": 1.817,
"eval_validation_privacy_sources_steps_per_second": 0.911,
"step": 4301
},
{
"epoch": 0.36,
"eval_validation_agenda_digitale_loss": 1.36287522315979,
"eval_validation_agenda_digitale_runtime": 187.0978,
"eval_validation_agenda_digitale_samples_per_second": 1.817,
"eval_validation_agenda_digitale_steps_per_second": 0.909,
"step": 4301
},
{
"epoch": 0.36,
"eval_validation_leggepertutti_loss": 1.225502371788025,
"eval_validation_leggepertutti_runtime": 35.7725,
"eval_validation_leggepertutti_samples_per_second": 1.817,
"eval_validation_leggepertutti_steps_per_second": 0.922,
"step": 4301
},
{
"epoch": 0.36,
"eval_validation_altalexprivacy_loss": 1.1899324655532837,
"eval_validation_altalexprivacy_runtime": 58.3269,
"eval_validation_altalexprivacy_samples_per_second": 1.817,
"eval_validation_altalexprivacy_steps_per_second": 0.909,
"step": 4301
},
{
"epoch": 0.36,
"eval_validation_mc4_it_loss": 1.7518540620803833,
"eval_validation_mc4_it_runtime": 78.1395,
"eval_validation_mc4_it_samples_per_second": 1.817,
"eval_validation_mc4_it_steps_per_second": 0.909,
"step": 4301
},
{
"epoch": 0.36,
"eval_validation_dolma_loss": 1.9364405870437622,
"eval_validation_dolma_runtime": 53.9366,
"eval_validation_dolma_samples_per_second": 1.817,
"eval_validation_dolma_steps_per_second": 0.908,
"step": 4301
},
{
"epoch": 0.38,
"grad_norm": 3.4375,
"learning_rate": 1e-05,
"loss": 1.5349,
"step": 4488
},
{
"epoch": 0.38,
"eval_validation_privacy_sources_loss": 0.8089282512664795,
"eval_validation_privacy_sources_runtime": 210.6735,
"eval_validation_privacy_sources_samples_per_second": 1.818,
"eval_validation_privacy_sources_steps_per_second": 0.911,
"step": 4488
},
{
"epoch": 0.38,
"eval_validation_agenda_digitale_loss": 1.3604696989059448,
"eval_validation_agenda_digitale_runtime": 187.0247,
"eval_validation_agenda_digitale_samples_per_second": 1.818,
"eval_validation_agenda_digitale_steps_per_second": 0.909,
"step": 4488
},
{
"epoch": 0.38,
"eval_validation_leggepertutti_loss": 1.2236857414245605,
"eval_validation_leggepertutti_runtime": 35.775,
"eval_validation_leggepertutti_samples_per_second": 1.817,
"eval_validation_leggepertutti_steps_per_second": 0.922,
"step": 4488
},
{
"epoch": 0.38,
"eval_validation_altalexprivacy_loss": 1.1892808675765991,
"eval_validation_altalexprivacy_runtime": 58.2983,
"eval_validation_altalexprivacy_samples_per_second": 1.818,
"eval_validation_altalexprivacy_steps_per_second": 0.909,
"step": 4488
},
{
"epoch": 0.38,
"eval_validation_mc4_it_loss": 1.7490640878677368,
"eval_validation_mc4_it_runtime": 78.1304,
"eval_validation_mc4_it_samples_per_second": 1.817,
"eval_validation_mc4_it_steps_per_second": 0.909,
"step": 4488
},
{
"epoch": 0.38,
"eval_validation_dolma_loss": 1.9361358880996704,
"eval_validation_dolma_runtime": 53.9258,
"eval_validation_dolma_samples_per_second": 1.817,
"eval_validation_dolma_steps_per_second": 0.909,
"step": 4488
},
{
"epoch": 0.39,
"grad_norm": 3.265625,
"learning_rate": 1e-05,
"loss": 1.538,
"step": 4675
},
{
"epoch": 0.39,
"eval_validation_privacy_sources_loss": 0.8068882822990417,
"eval_validation_privacy_sources_runtime": 210.6987,
"eval_validation_privacy_sources_samples_per_second": 1.818,
"eval_validation_privacy_sources_steps_per_second": 0.911,
"step": 4675
},
{
"epoch": 0.39,
"eval_validation_agenda_digitale_loss": 1.357992172241211,
"eval_validation_agenda_digitale_runtime": 187.0601,
"eval_validation_agenda_digitale_samples_per_second": 1.818,
"eval_validation_agenda_digitale_steps_per_second": 0.909,
"step": 4675
},
{
"epoch": 0.39,
"eval_validation_leggepertutti_loss": 1.2222189903259277,
"eval_validation_leggepertutti_runtime": 35.7685,
"eval_validation_leggepertutti_samples_per_second": 1.817,
"eval_validation_leggepertutti_steps_per_second": 0.923,
"step": 4675
},
{
"epoch": 0.39,
"eval_validation_altalexprivacy_loss": 1.1868293285369873,
"eval_validation_altalexprivacy_runtime": 58.3028,
"eval_validation_altalexprivacy_samples_per_second": 1.818,
"eval_validation_altalexprivacy_steps_per_second": 0.909,
"step": 4675
},
{
"epoch": 0.39,
"eval_validation_mc4_it_loss": 1.747467279434204,
"eval_validation_mc4_it_runtime": 78.0822,
"eval_validation_mc4_it_samples_per_second": 1.819,
"eval_validation_mc4_it_steps_per_second": 0.909,
"step": 4675
},
{
"epoch": 0.39,
"eval_validation_dolma_loss": 1.9376739263534546,
"eval_validation_dolma_runtime": 53.9131,
"eval_validation_dolma_samples_per_second": 1.818,
"eval_validation_dolma_steps_per_second": 0.909,
"step": 4675
},
{
"epoch": 0.41,
"grad_norm": 4.5625,
"learning_rate": 1e-05,
"loss": 1.5217,
"step": 4862
},
{
"epoch": 0.41,
"eval_validation_privacy_sources_loss": 0.8034628033638,
"eval_validation_privacy_sources_runtime": 210.664,
"eval_validation_privacy_sources_samples_per_second": 1.818,
"eval_validation_privacy_sources_steps_per_second": 0.911,
"step": 4862
},
{
"epoch": 0.41,
"eval_validation_agenda_digitale_loss": 1.3556355237960815,
"eval_validation_agenda_digitale_runtime": 187.0481,
"eval_validation_agenda_digitale_samples_per_second": 1.818,
"eval_validation_agenda_digitale_steps_per_second": 0.909,
"step": 4862
},
{
"epoch": 0.41,
"eval_validation_leggepertutti_loss": 1.2238881587982178,
"eval_validation_leggepertutti_runtime": 35.77,
"eval_validation_leggepertutti_samples_per_second": 1.817,
"eval_validation_leggepertutti_steps_per_second": 0.923,
"step": 4862
},
{
"epoch": 0.41,
"eval_validation_altalexprivacy_loss": 1.1821054220199585,
"eval_validation_altalexprivacy_runtime": 58.3131,
"eval_validation_altalexprivacy_samples_per_second": 1.818,
"eval_validation_altalexprivacy_steps_per_second": 0.909,
"step": 4862
},
{
"epoch": 0.41,
"eval_validation_mc4_it_loss": 1.7454167604446411,
"eval_validation_mc4_it_runtime": 78.136,
"eval_validation_mc4_it_samples_per_second": 1.817,
"eval_validation_mc4_it_steps_per_second": 0.909,
"step": 4862
},
{
"epoch": 0.41,
"eval_validation_dolma_loss": 1.9336557388305664,
"eval_validation_dolma_runtime": 53.931,
"eval_validation_dolma_samples_per_second": 1.817,
"eval_validation_dolma_steps_per_second": 0.909,
"step": 4862
},
{
"epoch": 0.43,
"grad_norm": 3.25,
"learning_rate": 1e-05,
"loss": 1.5231,
"step": 5049
},
{
"epoch": 0.43,
"eval_validation_privacy_sources_loss": 0.8018712997436523,
"eval_validation_privacy_sources_runtime": 210.6286,
"eval_validation_privacy_sources_samples_per_second": 1.818,
"eval_validation_privacy_sources_steps_per_second": 0.912,
"step": 5049
},
{
"epoch": 0.43,
"eval_validation_agenda_digitale_loss": 1.3536696434020996,
"eval_validation_agenda_digitale_runtime": 186.962,
"eval_validation_agenda_digitale_samples_per_second": 1.819,
"eval_validation_agenda_digitale_steps_per_second": 0.909,
"step": 5049
},
{
"epoch": 0.43,
"eval_validation_leggepertutti_loss": 1.219115138053894,
"eval_validation_leggepertutti_runtime": 35.7552,
"eval_validation_leggepertutti_samples_per_second": 1.818,
"eval_validation_leggepertutti_steps_per_second": 0.923,
"step": 5049
},
{
"epoch": 0.43,
"eval_validation_altalexprivacy_loss": 1.1803195476531982,
"eval_validation_altalexprivacy_runtime": 58.254,
"eval_validation_altalexprivacy_samples_per_second": 1.82,
"eval_validation_altalexprivacy_steps_per_second": 0.91,
"step": 5049
},
{
"epoch": 0.43,
"eval_validation_mc4_it_loss": 1.7427842617034912,
"eval_validation_mc4_it_runtime": 78.0819,
"eval_validation_mc4_it_samples_per_second": 1.819,
"eval_validation_mc4_it_steps_per_second": 0.909,
"step": 5049
},
{
"epoch": 0.43,
"eval_validation_dolma_loss": 1.932908058166504,
"eval_validation_dolma_runtime": 53.9066,
"eval_validation_dolma_samples_per_second": 1.818,
"eval_validation_dolma_steps_per_second": 0.909,
"step": 5049
},
{
"epoch": 0.44,
"grad_norm": 3.34375,
"learning_rate": 1e-05,
"loss": 1.535,
"step": 5236
},
{
"epoch": 0.44,
"eval_validation_privacy_sources_loss": 0.7990080714225769,
"eval_validation_privacy_sources_runtime": 210.6936,
"eval_validation_privacy_sources_samples_per_second": 1.818,
"eval_validation_privacy_sources_steps_per_second": 0.911,
"step": 5236
},
{
"epoch": 0.44,
"eval_validation_agenda_digitale_loss": 1.3512707948684692,
"eval_validation_agenda_digitale_runtime": 187.0661,
"eval_validation_agenda_digitale_samples_per_second": 1.818,
"eval_validation_agenda_digitale_steps_per_second": 0.909,
"step": 5236
},
{
"epoch": 0.44,
"eval_validation_leggepertutti_loss": 1.2183226346969604,
"eval_validation_leggepertutti_runtime": 35.7756,
"eval_validation_leggepertutti_samples_per_second": 1.817,
"eval_validation_leggepertutti_steps_per_second": 0.922,
"step": 5236
},
{
"epoch": 0.44,
"eval_validation_altalexprivacy_loss": 1.1762974262237549,
"eval_validation_altalexprivacy_runtime": 58.3182,
"eval_validation_altalexprivacy_samples_per_second": 1.818,
"eval_validation_altalexprivacy_steps_per_second": 0.909,
"step": 5236
},
{
"epoch": 0.44,
"eval_validation_mc4_it_loss": 1.7408561706542969,
"eval_validation_mc4_it_runtime": 78.1139,
"eval_validation_mc4_it_samples_per_second": 1.818,
"eval_validation_mc4_it_steps_per_second": 0.909,
"step": 5236
},
{
"epoch": 0.44,
"eval_validation_dolma_loss": 1.9324188232421875,
"eval_validation_dolma_runtime": 53.9296,
"eval_validation_dolma_samples_per_second": 1.817,
"eval_validation_dolma_steps_per_second": 0.909,
"step": 5236
},
{
"epoch": 0.46,
"grad_norm": 3.40625,
"learning_rate": 1e-05,
"loss": 1.5195,
"step": 5423
},
{
"epoch": 0.46,
"eval_validation_privacy_sources_loss": 0.7970147728919983,
"eval_validation_privacy_sources_runtime": 210.7209,
"eval_validation_privacy_sources_samples_per_second": 1.818,
"eval_validation_privacy_sources_steps_per_second": 0.911,
"step": 5423
},
{
"epoch": 0.46,
"eval_validation_agenda_digitale_loss": 1.350595235824585,
"eval_validation_agenda_digitale_runtime": 187.0929,
"eval_validation_agenda_digitale_samples_per_second": 1.817,
"eval_validation_agenda_digitale_steps_per_second": 0.909,
"step": 5423
},
{
"epoch": 0.46,
"eval_validation_leggepertutti_loss": 1.2196227312088013,
"eval_validation_leggepertutti_runtime": 35.7764,
"eval_validation_leggepertutti_samples_per_second": 1.817,
"eval_validation_leggepertutti_steps_per_second": 0.922,
"step": 5423
},
{
"epoch": 0.46,
"eval_validation_altalexprivacy_loss": 1.176468014717102,
"eval_validation_altalexprivacy_runtime": 58.3399,
"eval_validation_altalexprivacy_samples_per_second": 1.817,
"eval_validation_altalexprivacy_steps_per_second": 0.908,
"step": 5423
},
{
"epoch": 0.46,
"eval_validation_mc4_it_loss": 1.7395856380462646,
"eval_validation_mc4_it_runtime": 78.1494,
"eval_validation_mc4_it_samples_per_second": 1.817,
"eval_validation_mc4_it_steps_per_second": 0.909,
"step": 5423
},
{
"epoch": 0.46,
"eval_validation_dolma_loss": 1.9318716526031494,
"eval_validation_dolma_runtime": 53.9467,
"eval_validation_dolma_samples_per_second": 1.817,
"eval_validation_dolma_steps_per_second": 0.908,
"step": 5423
},
{
"epoch": 0.47,
"grad_norm": 3.484375,
"learning_rate": 1e-05,
"loss": 1.5356,
"step": 5610
},
{
"epoch": 0.47,
"eval_validation_privacy_sources_loss": 0.7967193126678467,
"eval_validation_privacy_sources_runtime": 210.7692,
"eval_validation_privacy_sources_samples_per_second": 1.817,
"eval_validation_privacy_sources_steps_per_second": 0.911,
"step": 5610
},
{
"epoch": 0.47,
"eval_validation_agenda_digitale_loss": 1.3498324155807495,
"eval_validation_agenda_digitale_runtime": 187.1217,
"eval_validation_agenda_digitale_samples_per_second": 1.817,
"eval_validation_agenda_digitale_steps_per_second": 0.908,
"step": 5610
},
{
"epoch": 0.47,
"eval_validation_leggepertutti_loss": 1.2213656902313232,
"eval_validation_leggepertutti_runtime": 35.7931,
"eval_validation_leggepertutti_samples_per_second": 1.816,
"eval_validation_leggepertutti_steps_per_second": 0.922,
"step": 5610
},
{
"epoch": 0.47,
"eval_validation_altalexprivacy_loss": 1.1725049018859863,
"eval_validation_altalexprivacy_runtime": 58.34,
"eval_validation_altalexprivacy_samples_per_second": 1.817,
"eval_validation_altalexprivacy_steps_per_second": 0.908,
"step": 5610
},
{
"epoch": 0.47,
"eval_validation_mc4_it_loss": 1.7369210720062256,
"eval_validation_mc4_it_runtime": 78.1677,
"eval_validation_mc4_it_samples_per_second": 1.817,
"eval_validation_mc4_it_steps_per_second": 0.908,
"step": 5610
},
{
"epoch": 0.47,
"eval_validation_dolma_loss": 1.931601643562317,
"eval_validation_dolma_runtime": 53.9523,
"eval_validation_dolma_samples_per_second": 1.816,
"eval_validation_dolma_steps_per_second": 0.908,
"step": 5610
},
{
"epoch": 0.49,
"grad_norm": 3.25,
"learning_rate": 1e-05,
"loss": 1.5248,
"step": 5797
},
{
"epoch": 0.49,
"eval_validation_privacy_sources_loss": 0.7927350997924805,
"eval_validation_privacy_sources_runtime": 210.7499,
"eval_validation_privacy_sources_samples_per_second": 1.817,
"eval_validation_privacy_sources_steps_per_second": 0.911,
"step": 5797
},
{
"epoch": 0.49,
"eval_validation_agenda_digitale_loss": 1.3474359512329102,
"eval_validation_agenda_digitale_runtime": 187.1385,
"eval_validation_agenda_digitale_samples_per_second": 1.817,
"eval_validation_agenda_digitale_steps_per_second": 0.908,
"step": 5797
},
{
"epoch": 0.49,
"eval_validation_leggepertutti_loss": 1.2162562608718872,
"eval_validation_leggepertutti_runtime": 35.7966,
"eval_validation_leggepertutti_samples_per_second": 1.816,
"eval_validation_leggepertutti_steps_per_second": 0.922,
"step": 5797
},
{
"epoch": 0.49,
"eval_validation_altalexprivacy_loss": 1.1715296506881714,
"eval_validation_altalexprivacy_runtime": 58.3612,
"eval_validation_altalexprivacy_samples_per_second": 1.816,
"eval_validation_altalexprivacy_steps_per_second": 0.908,
"step": 5797
},
{
"epoch": 0.49,
"eval_validation_mc4_it_loss": 1.7344577312469482,
"eval_validation_mc4_it_runtime": 78.1843,
"eval_validation_mc4_it_samples_per_second": 1.816,
"eval_validation_mc4_it_steps_per_second": 0.908,
"step": 5797
},
{
"epoch": 0.49,
"eval_validation_dolma_loss": 1.9310978651046753,
"eval_validation_dolma_runtime": 53.9513,
"eval_validation_dolma_samples_per_second": 1.816,
"eval_validation_dolma_steps_per_second": 0.908,
"step": 5797
},
{
"epoch": 0.5,
"grad_norm": 3.390625,
"learning_rate": 1e-05,
"loss": 1.5147,
"step": 5984
},
{
"epoch": 0.5,
"eval_validation_privacy_sources_loss": 0.7904417514801025,
"eval_validation_privacy_sources_runtime": 210.6793,
"eval_validation_privacy_sources_samples_per_second": 1.818,
"eval_validation_privacy_sources_steps_per_second": 0.911,
"step": 5984
},
{
"epoch": 0.5,
"eval_validation_agenda_digitale_loss": 1.3463512659072876,
"eval_validation_agenda_digitale_runtime": 187.0767,
"eval_validation_agenda_digitale_samples_per_second": 1.817,
"eval_validation_agenda_digitale_steps_per_second": 0.909,
"step": 5984
},
{
"epoch": 0.5,
"eval_validation_leggepertutti_loss": 1.2104567289352417,
"eval_validation_leggepertutti_runtime": 35.7676,
"eval_validation_leggepertutti_samples_per_second": 1.817,
"eval_validation_leggepertutti_steps_per_second": 0.923,
"step": 5984
},
{
"epoch": 0.5,
"eval_validation_altalexprivacy_loss": 1.1698046922683716,
"eval_validation_altalexprivacy_runtime": 58.3216,
"eval_validation_altalexprivacy_samples_per_second": 1.818,
"eval_validation_altalexprivacy_steps_per_second": 0.909,
"step": 5984
},
{
"epoch": 0.5,
"eval_validation_mc4_it_loss": 1.733211874961853,
"eval_validation_mc4_it_runtime": 78.1157,
"eval_validation_mc4_it_samples_per_second": 1.818,
"eval_validation_mc4_it_steps_per_second": 0.909,
"step": 5984
},
{
"epoch": 0.5,
"eval_validation_dolma_loss": 1.9291273355484009,
"eval_validation_dolma_runtime": 53.9368,
"eval_validation_dolma_samples_per_second": 1.817,
"eval_validation_dolma_steps_per_second": 0.908,
"step": 5984
},
{
"epoch": 0.52,
"grad_norm": 3.15625,
"learning_rate": 1e-05,
"loss": 1.5307,
"step": 6171
},
{
"epoch": 0.52,
"eval_validation_privacy_sources_loss": 0.7893861532211304,
"eval_validation_privacy_sources_runtime": 210.7488,
"eval_validation_privacy_sources_samples_per_second": 1.817,
"eval_validation_privacy_sources_steps_per_second": 0.911,
"step": 6171
},
{
"epoch": 0.52,
"eval_validation_agenda_digitale_loss": 1.3451480865478516,
"eval_validation_agenda_digitale_runtime": 187.0845,
"eval_validation_agenda_digitale_samples_per_second": 1.817,
"eval_validation_agenda_digitale_steps_per_second": 0.909,
"step": 6171
},
{
"epoch": 0.52,
"eval_validation_leggepertutti_loss": 1.2113819122314453,
"eval_validation_leggepertutti_runtime": 35.7727,
"eval_validation_leggepertutti_samples_per_second": 1.817,
"eval_validation_leggepertutti_steps_per_second": 0.922,
"step": 6171
},
{
"epoch": 0.52,
"eval_validation_altalexprivacy_loss": 1.1663610935211182,
"eval_validation_altalexprivacy_runtime": 58.3307,
"eval_validation_altalexprivacy_samples_per_second": 1.817,
"eval_validation_altalexprivacy_steps_per_second": 0.909,
"step": 6171
},
{
"epoch": 0.52,
"eval_validation_mc4_it_loss": 1.7309118509292603,
"eval_validation_mc4_it_runtime": 78.1316,
"eval_validation_mc4_it_samples_per_second": 1.817,
"eval_validation_mc4_it_steps_per_second": 0.909,
"step": 6171
},
{
"epoch": 0.52,
"eval_validation_dolma_loss": 1.9281673431396484,
"eval_validation_dolma_runtime": 53.9207,
"eval_validation_dolma_samples_per_second": 1.817,
"eval_validation_dolma_steps_per_second": 0.909,
"step": 6171
},
{
"epoch": 0.54,
"grad_norm": 3.078125,
"learning_rate": 1e-05,
"loss": 1.5203,
"step": 6358
},
{
"epoch": 0.54,
"eval_validation_privacy_sources_loss": 0.7874845266342163,
"eval_validation_privacy_sources_runtime": 210.6148,
"eval_validation_privacy_sources_samples_per_second": 1.818,
"eval_validation_privacy_sources_steps_per_second": 0.912,
"step": 6358
},
{
"epoch": 0.54,
"eval_validation_agenda_digitale_loss": 1.3434906005859375,
"eval_validation_agenda_digitale_runtime": 186.9358,
"eval_validation_agenda_digitale_samples_per_second": 1.819,
"eval_validation_agenda_digitale_steps_per_second": 0.909,
"step": 6358
},
{
"epoch": 0.54,
"eval_validation_leggepertutti_loss": 1.2103608846664429,
"eval_validation_leggepertutti_runtime": 35.7522,
"eval_validation_leggepertutti_samples_per_second": 1.818,
"eval_validation_leggepertutti_steps_per_second": 0.923,
"step": 6358
},
{
"epoch": 0.54,
"eval_validation_altalexprivacy_loss": 1.164802074432373,
"eval_validation_altalexprivacy_runtime": 58.2954,
"eval_validation_altalexprivacy_samples_per_second": 1.818,
"eval_validation_altalexprivacy_steps_per_second": 0.909,
"step": 6358
},
{
"epoch": 0.54,
"eval_validation_mc4_it_loss": 1.7286055088043213,
"eval_validation_mc4_it_runtime": 78.0929,
"eval_validation_mc4_it_samples_per_second": 1.818,
"eval_validation_mc4_it_steps_per_second": 0.909,
"step": 6358
},
{
"epoch": 0.54,
"eval_validation_dolma_loss": 1.927778720855713,
"eval_validation_dolma_runtime": 53.9081,
"eval_validation_dolma_samples_per_second": 1.818,
"eval_validation_dolma_steps_per_second": 0.909,
"step": 6358
},
{
"epoch": 0.55,
"grad_norm": 3.59375,
"learning_rate": 1e-05,
"loss": 1.5257,
"step": 6545
},
{
"epoch": 0.55,
"eval_validation_privacy_sources_loss": 0.785801887512207,
"eval_validation_privacy_sources_runtime": 210.655,
"eval_validation_privacy_sources_samples_per_second": 1.818,
"eval_validation_privacy_sources_steps_per_second": 0.911,
"step": 6545
},
{
"epoch": 0.55,
"eval_validation_agenda_digitale_loss": 1.341984510421753,
"eval_validation_agenda_digitale_runtime": 186.9769,
"eval_validation_agenda_digitale_samples_per_second": 1.818,
"eval_validation_agenda_digitale_steps_per_second": 0.909,
"step": 6545
},
{
"epoch": 0.55,
"eval_validation_leggepertutti_loss": 1.208351492881775,
"eval_validation_leggepertutti_runtime": 35.7494,
"eval_validation_leggepertutti_samples_per_second": 1.818,
"eval_validation_leggepertutti_steps_per_second": 0.923,
"step": 6545
},
{
"epoch": 0.55,
"eval_validation_altalexprivacy_loss": 1.1633599996566772,
"eval_validation_altalexprivacy_runtime": 58.2944,
"eval_validation_altalexprivacy_samples_per_second": 1.818,
"eval_validation_altalexprivacy_steps_per_second": 0.909,
"step": 6545
},
{
"epoch": 0.55,
"eval_validation_mc4_it_loss": 1.726757526397705,
"eval_validation_mc4_it_runtime": 78.0983,
"eval_validation_mc4_it_samples_per_second": 1.818,
"eval_validation_mc4_it_steps_per_second": 0.909,
"step": 6545
},
{
"epoch": 0.55,
"eval_validation_dolma_loss": 1.9275562763214111,
"eval_validation_dolma_runtime": 53.8914,
"eval_validation_dolma_samples_per_second": 1.818,
"eval_validation_dolma_steps_per_second": 0.909,
"step": 6545
},
{
"epoch": 0.57,
"grad_norm": 3.234375,
"learning_rate": 1e-05,
"loss": 1.5019,
"step": 6732
},
{
"epoch": 0.57,
"eval_validation_privacy_sources_loss": 0.7848771214485168,
"eval_validation_privacy_sources_runtime": 210.6636,
"eval_validation_privacy_sources_samples_per_second": 1.818,
"eval_validation_privacy_sources_steps_per_second": 0.911,
"step": 6732
},
{
"epoch": 0.57,
"eval_validation_agenda_digitale_loss": 1.3411486148834229,
"eval_validation_agenda_digitale_runtime": 186.9976,
"eval_validation_agenda_digitale_samples_per_second": 1.818,
"eval_validation_agenda_digitale_steps_per_second": 0.909,
"step": 6732
},
{
"epoch": 0.57,
"eval_validation_leggepertutti_loss": 1.2077082395553589,
"eval_validation_leggepertutti_runtime": 35.7665,
"eval_validation_leggepertutti_samples_per_second": 1.817,
"eval_validation_leggepertutti_steps_per_second": 0.923,
"step": 6732
},
{
"epoch": 0.57,
"eval_validation_altalexprivacy_loss": 1.1618914604187012,
"eval_validation_altalexprivacy_runtime": 58.33,
"eval_validation_altalexprivacy_samples_per_second": 1.817,
"eval_validation_altalexprivacy_steps_per_second": 0.909,
"step": 6732
},
{
"epoch": 0.57,
"eval_validation_mc4_it_loss": 1.725679874420166,
"eval_validation_mc4_it_runtime": 78.1276,
"eval_validation_mc4_it_samples_per_second": 1.818,
"eval_validation_mc4_it_steps_per_second": 0.909,
"step": 6732
},
{
"epoch": 0.57,
"eval_validation_dolma_loss": 1.9271798133850098,
"eval_validation_dolma_runtime": 53.9117,
"eval_validation_dolma_samples_per_second": 1.818,
"eval_validation_dolma_steps_per_second": 0.909,
"step": 6732
},
{
"epoch": 0.58,
"grad_norm": 3.15625,
"learning_rate": 1e-05,
"loss": 1.5152,
"step": 6919
},
{
"epoch": 0.58,
"eval_validation_privacy_sources_loss": 0.782995879650116,
"eval_validation_privacy_sources_runtime": 210.7422,
"eval_validation_privacy_sources_samples_per_second": 1.817,
"eval_validation_privacy_sources_steps_per_second": 0.911,
"step": 6919
},
{
"epoch": 0.58,
"eval_validation_agenda_digitale_loss": 1.3396461009979248,
"eval_validation_agenda_digitale_runtime": 187.0789,
"eval_validation_agenda_digitale_samples_per_second": 1.817,
"eval_validation_agenda_digitale_steps_per_second": 0.909,
"step": 6919
},
{
"epoch": 0.58,
"eval_validation_leggepertutti_loss": 1.2054615020751953,
"eval_validation_leggepertutti_runtime": 35.7709,
"eval_validation_leggepertutti_samples_per_second": 1.817,
"eval_validation_leggepertutti_steps_per_second": 0.923,
"step": 6919
},
{
"epoch": 0.58,
"eval_validation_altalexprivacy_loss": 1.1602697372436523,
"eval_validation_altalexprivacy_runtime": 58.3175,
"eval_validation_altalexprivacy_samples_per_second": 1.818,
"eval_validation_altalexprivacy_steps_per_second": 0.909,
"step": 6919
},
{
"epoch": 0.58,
"eval_validation_mc4_it_loss": 1.724775791168213,
"eval_validation_mc4_it_runtime": 78.0214,
"eval_validation_mc4_it_samples_per_second": 1.82,
"eval_validation_mc4_it_steps_per_second": 0.91,
"step": 6919
},
{
"epoch": 0.58,
"eval_validation_dolma_loss": 1.9262704849243164,
"eval_validation_dolma_runtime": 53.8374,
"eval_validation_dolma_samples_per_second": 1.82,
"eval_validation_dolma_steps_per_second": 0.91,
"step": 6919
},
{
"epoch": 0.6,
"grad_norm": 3.28125,
"learning_rate": 1e-05,
"loss": 1.5167,
"step": 7106
},
{
"epoch": 0.6,
"eval_validation_privacy_sources_loss": 0.7818235158920288,
"eval_validation_privacy_sources_runtime": 210.7745,
"eval_validation_privacy_sources_samples_per_second": 1.817,
"eval_validation_privacy_sources_steps_per_second": 0.911,
"step": 7106
},
{
"epoch": 0.6,
"eval_validation_agenda_digitale_loss": 1.3387728929519653,
"eval_validation_agenda_digitale_runtime": 187.1722,
"eval_validation_agenda_digitale_samples_per_second": 1.817,
"eval_validation_agenda_digitale_steps_per_second": 0.908,
"step": 7106
},
{
"epoch": 0.6,
"eval_validation_leggepertutti_loss": 1.2060880661010742,
"eval_validation_leggepertutti_runtime": 35.7758,
"eval_validation_leggepertutti_samples_per_second": 1.817,
"eval_validation_leggepertutti_steps_per_second": 0.922,
"step": 7106
},
{
"epoch": 0.6,
"eval_validation_altalexprivacy_loss": 1.158158302307129,
"eval_validation_altalexprivacy_runtime": 58.3272,
"eval_validation_altalexprivacy_samples_per_second": 1.817,
"eval_validation_altalexprivacy_steps_per_second": 0.909,
"step": 7106
},
{
"epoch": 0.6,
"eval_validation_mc4_it_loss": 1.7227188348770142,
"eval_validation_mc4_it_runtime": 78.0893,
"eval_validation_mc4_it_samples_per_second": 1.818,
"eval_validation_mc4_it_steps_per_second": 0.909,
"step": 7106
},
{
"epoch": 0.6,
"eval_validation_dolma_loss": 1.925492763519287,
"eval_validation_dolma_runtime": 53.8538,
"eval_validation_dolma_samples_per_second": 1.82,
"eval_validation_dolma_steps_per_second": 0.91,
"step": 7106
},
{
"epoch": 0.61,
"grad_norm": 3.390625,
"learning_rate": 1e-05,
"loss": 1.4819,
"step": 7293
},
{
"epoch": 0.61,
"eval_validation_privacy_sources_loss": 0.7804288268089294,
"eval_validation_privacy_sources_runtime": 210.7884,
"eval_validation_privacy_sources_samples_per_second": 1.817,
"eval_validation_privacy_sources_steps_per_second": 0.911,
"step": 7293
},
{
"epoch": 0.61,
"eval_validation_agenda_digitale_loss": 1.337357521057129,
"eval_validation_agenda_digitale_runtime": 187.1395,
"eval_validation_agenda_digitale_samples_per_second": 1.817,
"eval_validation_agenda_digitale_steps_per_second": 0.908,
"step": 7293
},
{
"epoch": 0.61,
"eval_validation_leggepertutti_loss": 1.205103874206543,
"eval_validation_leggepertutti_runtime": 35.7974,
"eval_validation_leggepertutti_samples_per_second": 1.816,
"eval_validation_leggepertutti_steps_per_second": 0.922,
"step": 7293
},
{
"epoch": 0.61,
"eval_validation_altalexprivacy_loss": 1.155694603919983,
"eval_validation_altalexprivacy_runtime": 58.3571,
"eval_validation_altalexprivacy_samples_per_second": 1.816,
"eval_validation_altalexprivacy_steps_per_second": 0.908,
"step": 7293
},
{
"epoch": 0.61,
"eval_validation_mc4_it_loss": 1.7214677333831787,
"eval_validation_mc4_it_runtime": 78.1458,
"eval_validation_mc4_it_samples_per_second": 1.817,
"eval_validation_mc4_it_steps_per_second": 0.909,
"step": 7293
},
{
"epoch": 0.61,
"eval_validation_dolma_loss": 1.926133155822754,
"eval_validation_dolma_runtime": 53.9449,
"eval_validation_dolma_samples_per_second": 1.817,
"eval_validation_dolma_steps_per_second": 0.908,
"step": 7293
},
{
"epoch": 0.63,
"grad_norm": 3.34375,
"learning_rate": 1e-05,
"loss": 1.5164,
"step": 7480
},
{
"epoch": 0.63,
"eval_validation_privacy_sources_loss": 0.7794305682182312,
"eval_validation_privacy_sources_runtime": 210.7859,
"eval_validation_privacy_sources_samples_per_second": 1.817,
"eval_validation_privacy_sources_steps_per_second": 0.911,
"step": 7480
},
{
"epoch": 0.63,
"eval_validation_agenda_digitale_loss": 1.3364444971084595,
"eval_validation_agenda_digitale_runtime": 187.1772,
"eval_validation_agenda_digitale_samples_per_second": 1.816,
"eval_validation_agenda_digitale_steps_per_second": 0.908,
"step": 7480
},
{
"epoch": 0.63,
"eval_validation_leggepertutti_loss": 1.2025309801101685,
"eval_validation_leggepertutti_runtime": 35.7941,
"eval_validation_leggepertutti_samples_per_second": 1.816,
"eval_validation_leggepertutti_steps_per_second": 0.922,
"step": 7480
},
{
"epoch": 0.63,
"eval_validation_altalexprivacy_loss": 1.1547417640686035,
"eval_validation_altalexprivacy_runtime": 58.3512,
"eval_validation_altalexprivacy_samples_per_second": 1.817,
"eval_validation_altalexprivacy_steps_per_second": 0.908,
"step": 7480
},
{
"epoch": 0.63,
"eval_validation_mc4_it_loss": 1.7210993766784668,
"eval_validation_mc4_it_runtime": 78.1723,
"eval_validation_mc4_it_samples_per_second": 1.816,
"eval_validation_mc4_it_steps_per_second": 0.908,
"step": 7480
},
{
"epoch": 0.63,
"eval_validation_dolma_loss": 1.925874948501587,
"eval_validation_dolma_runtime": 53.9674,
"eval_validation_dolma_samples_per_second": 1.816,
"eval_validation_dolma_steps_per_second": 0.908,
"step": 7480
},
{
"epoch": 0.65,
"grad_norm": 3.140625,
"learning_rate": 1e-05,
"loss": 1.5167,
"step": 7667
},
{
"epoch": 0.65,
"eval_validation_privacy_sources_loss": 0.7769683599472046,
"eval_validation_privacy_sources_runtime": 210.8943,
"eval_validation_privacy_sources_samples_per_second": 1.816,
"eval_validation_privacy_sources_steps_per_second": 0.91,
"step": 7667
},
{
"epoch": 0.65,
"eval_validation_agenda_digitale_loss": 1.3346428871154785,
"eval_validation_agenda_digitale_runtime": 187.2224,
"eval_validation_agenda_digitale_samples_per_second": 1.816,
"eval_validation_agenda_digitale_steps_per_second": 0.908,
"step": 7667
},
{
"epoch": 0.65,
"eval_validation_leggepertutti_loss": 1.2033891677856445,
"eval_validation_leggepertutti_runtime": 35.8133,
"eval_validation_leggepertutti_samples_per_second": 1.815,
"eval_validation_leggepertutti_steps_per_second": 0.921,
"step": 7667
},
{
"epoch": 0.65,
"eval_validation_altalexprivacy_loss": 1.1526622772216797,
"eval_validation_altalexprivacy_runtime": 58.3616,
"eval_validation_altalexprivacy_samples_per_second": 1.816,
"eval_validation_altalexprivacy_steps_per_second": 0.908,
"step": 7667
},
{
"epoch": 0.65,
"eval_validation_mc4_it_loss": 1.7187703847885132,
"eval_validation_mc4_it_runtime": 78.1967,
"eval_validation_mc4_it_samples_per_second": 1.816,
"eval_validation_mc4_it_steps_per_second": 0.908,
"step": 7667
},
{
"epoch": 0.65,
"eval_validation_dolma_loss": 1.9250844717025757,
"eval_validation_dolma_runtime": 53.9678,
"eval_validation_dolma_samples_per_second": 1.816,
"eval_validation_dolma_steps_per_second": 0.908,
"step": 7667
}
],
"logging_steps": 187,
"max_steps": 23720,
"num_input_tokens_seen": 0,
"num_train_epochs": 2,
"save_steps": 187,
"total_flos": 3.215568919621062e+19,
"train_batch_size": 1,
"trial_name": null,
"trial_params": null
}