LexLLMv0.0.3_ckpt3366 / trainer_state.json
andrealexroom's picture
Upload folder using huggingface_hub
44adce5 verified
{
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 0.41934127540294325,
"eval_steps": 187,
"global_step": 3366,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.00012458148407692906,
"grad_norm": 4.625,
"learning_rate": 1.2500000000000002e-07,
"loss": 1.3698,
"step": 1
},
{
"epoch": 0.023296737522385737,
"grad_norm": 1.3203125,
"learning_rate": 1e-05,
"loss": 1.3524,
"step": 187
},
{
"epoch": 0.023296737522385737,
"eval_validation_loss": 1.1260877847671509,
"eval_validation_runtime": 191.5792,
"eval_validation_samples_per_second": 1.295,
"eval_validation_steps_per_second": 1.295,
"step": 187
},
{
"epoch": 0.023296737522385737,
"eval_validation_privacy_sources_loss": 1.1260877847671509,
"eval_validation_privacy_sources_runtime": 191.505,
"eval_validation_privacy_sources_samples_per_second": 1.295,
"eval_validation_privacy_sources_steps_per_second": 1.295,
"step": 187
},
{
"epoch": 0.023296737522385737,
"eval_validation_agenda_digitale_loss": 1.621711015701294,
"eval_validation_agenda_digitale_runtime": 172.7434,
"eval_validation_agenda_digitale_samples_per_second": 1.297,
"eval_validation_agenda_digitale_steps_per_second": 1.297,
"step": 187
},
{
"epoch": 0.023296737522385737,
"eval_validation_legal_articles_loss": 1.499701738357544,
"eval_validation_legal_articles_runtime": 276.8245,
"eval_validation_legal_articles_samples_per_second": 1.297,
"eval_validation_legal_articles_steps_per_second": 1.297,
"step": 187
},
{
"epoch": 0.023296737522385737,
"eval_validation_leggepertutti_loss": 1.2891578674316406,
"eval_validation_leggepertutti_runtime": 30.8473,
"eval_validation_leggepertutti_samples_per_second": 1.297,
"eval_validation_leggepertutti_steps_per_second": 1.297,
"step": 187
},
{
"epoch": 0.046593475044771474,
"grad_norm": 1.734375,
"learning_rate": 1e-05,
"loss": 1.2846,
"step": 374
},
{
"epoch": 0.046593475044771474,
"eval_validation_loss": 1.094785213470459,
"eval_validation_runtime": 191.6632,
"eval_validation_samples_per_second": 1.294,
"eval_validation_steps_per_second": 1.294,
"step": 374
},
{
"epoch": 0.046593475044771474,
"eval_validation_privacy_sources_loss": 1.094785213470459,
"eval_validation_privacy_sources_runtime": 191.6482,
"eval_validation_privacy_sources_samples_per_second": 1.294,
"eval_validation_privacy_sources_steps_per_second": 1.294,
"step": 374
},
{
"epoch": 0.046593475044771474,
"eval_validation_agenda_digitale_loss": 1.6178028583526611,
"eval_validation_agenda_digitale_runtime": 173.068,
"eval_validation_agenda_digitale_samples_per_second": 1.294,
"eval_validation_agenda_digitale_steps_per_second": 1.294,
"step": 374
},
{
"epoch": 0.046593475044771474,
"eval_validation_legal_articles_loss": 1.4950900077819824,
"eval_validation_legal_articles_runtime": 277.5124,
"eval_validation_legal_articles_samples_per_second": 1.294,
"eval_validation_legal_articles_steps_per_second": 1.294,
"step": 374
},
{
"epoch": 0.046593475044771474,
"eval_validation_leggepertutti_loss": 1.2851760387420654,
"eval_validation_leggepertutti_runtime": 30.9036,
"eval_validation_leggepertutti_samples_per_second": 1.294,
"eval_validation_leggepertutti_steps_per_second": 1.294,
"step": 374
},
{
"epoch": 0.0698902125671572,
"grad_norm": 1.21875,
"learning_rate": 1e-05,
"loss": 1.2621,
"step": 561
},
{
"epoch": 0.0698902125671572,
"eval_validation_loss": 1.0743807554244995,
"eval_validation_runtime": 191.5602,
"eval_validation_samples_per_second": 1.295,
"eval_validation_steps_per_second": 1.295,
"step": 561
},
{
"epoch": 0.0698902125671572,
"eval_validation_privacy_sources_loss": 1.0743807554244995,
"eval_validation_privacy_sources_runtime": 191.5528,
"eval_validation_privacy_sources_samples_per_second": 1.295,
"eval_validation_privacy_sources_steps_per_second": 1.295,
"step": 561
},
{
"epoch": 0.0698902125671572,
"eval_validation_agenda_digitale_loss": 1.6161115169525146,
"eval_validation_agenda_digitale_runtime": 173.0656,
"eval_validation_agenda_digitale_samples_per_second": 1.294,
"eval_validation_agenda_digitale_steps_per_second": 1.294,
"step": 561
},
{
"epoch": 0.0698902125671572,
"eval_validation_legal_articles_loss": 1.4931106567382812,
"eval_validation_legal_articles_runtime": 277.3816,
"eval_validation_legal_articles_samples_per_second": 1.294,
"eval_validation_legal_articles_steps_per_second": 1.294,
"step": 561
},
{
"epoch": 0.0698902125671572,
"eval_validation_leggepertutti_loss": 1.2841417789459229,
"eval_validation_leggepertutti_runtime": 30.8957,
"eval_validation_leggepertutti_samples_per_second": 1.295,
"eval_validation_leggepertutti_steps_per_second": 1.295,
"step": 561
},
{
"epoch": 0.09318695008954295,
"grad_norm": 1.25,
"learning_rate": 1e-05,
"loss": 1.2494,
"step": 748
},
{
"epoch": 0.09318695008954295,
"eval_validation_loss": 1.0636643171310425,
"eval_validation_runtime": 191.5239,
"eval_validation_samples_per_second": 1.295,
"eval_validation_steps_per_second": 1.295,
"step": 748
},
{
"epoch": 0.09318695008954295,
"eval_validation_privacy_sources_loss": 1.0636643171310425,
"eval_validation_privacy_sources_runtime": 191.5331,
"eval_validation_privacy_sources_samples_per_second": 1.295,
"eval_validation_privacy_sources_steps_per_second": 1.295,
"step": 748
},
{
"epoch": 0.09318695008954295,
"eval_validation_agenda_digitale_loss": 1.613512635231018,
"eval_validation_agenda_digitale_runtime": 172.9978,
"eval_validation_agenda_digitale_samples_per_second": 1.295,
"eval_validation_agenda_digitale_steps_per_second": 1.295,
"step": 748
},
{
"epoch": 0.09318695008954295,
"eval_validation_legal_articles_loss": 1.4906712770462036,
"eval_validation_legal_articles_runtime": 277.2841,
"eval_validation_legal_articles_samples_per_second": 1.295,
"eval_validation_legal_articles_steps_per_second": 1.295,
"step": 748
},
{
"epoch": 0.09318695008954295,
"eval_validation_leggepertutti_loss": 1.2821543216705322,
"eval_validation_leggepertutti_runtime": 30.8792,
"eval_validation_leggepertutti_samples_per_second": 1.295,
"eval_validation_leggepertutti_steps_per_second": 1.295,
"step": 748
},
{
"epoch": 0.11648368761192868,
"grad_norm": 1.7265625,
"learning_rate": 1e-05,
"loss": 1.2266,
"step": 935
},
{
"epoch": 0.11648368761192868,
"eval_validation_loss": 1.0539060831069946,
"eval_validation_runtime": 191.4455,
"eval_validation_samples_per_second": 1.295,
"eval_validation_steps_per_second": 1.295,
"step": 935
},
{
"epoch": 0.11648368761192868,
"eval_validation_privacy_sources_loss": 1.0539060831069946,
"eval_validation_privacy_sources_runtime": 191.4432,
"eval_validation_privacy_sources_samples_per_second": 1.295,
"eval_validation_privacy_sources_steps_per_second": 1.295,
"step": 935
},
{
"epoch": 0.11648368761192868,
"eval_validation_agenda_digitale_loss": 1.6116217374801636,
"eval_validation_agenda_digitale_runtime": 172.9575,
"eval_validation_agenda_digitale_samples_per_second": 1.295,
"eval_validation_agenda_digitale_steps_per_second": 1.295,
"step": 935
},
{
"epoch": 0.11648368761192868,
"eval_validation_legal_articles_loss": 1.488964557647705,
"eval_validation_legal_articles_runtime": 277.1972,
"eval_validation_legal_articles_samples_per_second": 1.295,
"eval_validation_legal_articles_steps_per_second": 1.295,
"step": 935
},
{
"epoch": 0.11648368761192868,
"eval_validation_leggepertutti_loss": 1.2813745737075806,
"eval_validation_leggepertutti_runtime": 30.9233,
"eval_validation_leggepertutti_samples_per_second": 1.294,
"eval_validation_leggepertutti_steps_per_second": 1.294,
"step": 935
},
{
"epoch": 0.1397804251343144,
"grad_norm": 1.28125,
"learning_rate": 1e-05,
"loss": 1.2224,
"step": 1122
},
{
"epoch": 0.1397804251343144,
"eval_validation_loss": 1.0441668033599854,
"eval_validation_runtime": 191.4636,
"eval_validation_samples_per_second": 1.295,
"eval_validation_steps_per_second": 1.295,
"step": 1122
},
{
"epoch": 0.1397804251343144,
"eval_validation_privacy_sources_loss": 1.0441668033599854,
"eval_validation_privacy_sources_runtime": 191.473,
"eval_validation_privacy_sources_samples_per_second": 1.295,
"eval_validation_privacy_sources_steps_per_second": 1.295,
"step": 1122
},
{
"epoch": 0.1397804251343144,
"eval_validation_agenda_digitale_loss": 1.6095850467681885,
"eval_validation_agenda_digitale_runtime": 172.9659,
"eval_validation_agenda_digitale_samples_per_second": 1.295,
"eval_validation_agenda_digitale_steps_per_second": 1.295,
"step": 1122
},
{
"epoch": 0.1397804251343144,
"eval_validation_legal_articles_loss": 1.4869102239608765,
"eval_validation_legal_articles_runtime": 277.1724,
"eval_validation_legal_articles_samples_per_second": 1.295,
"eval_validation_legal_articles_steps_per_second": 1.295,
"step": 1122
},
{
"epoch": 0.1397804251343144,
"eval_validation_leggepertutti_loss": 1.2796441316604614,
"eval_validation_leggepertutti_runtime": 30.8854,
"eval_validation_leggepertutti_samples_per_second": 1.295,
"eval_validation_leggepertutti_steps_per_second": 1.295,
"step": 1122
},
{
"epoch": 0.16307716265670014,
"grad_norm": 1.1015625,
"learning_rate": 1e-05,
"loss": 1.2181,
"step": 1309
},
{
"epoch": 0.16307716265670014,
"eval_validation_loss": 1.0365145206451416,
"eval_validation_runtime": 191.4083,
"eval_validation_samples_per_second": 1.296,
"eval_validation_steps_per_second": 1.296,
"step": 1309
},
{
"epoch": 0.16307716265670014,
"eval_validation_privacy_sources_loss": 1.0365145206451416,
"eval_validation_privacy_sources_runtime": 191.4322,
"eval_validation_privacy_sources_samples_per_second": 1.295,
"eval_validation_privacy_sources_steps_per_second": 1.295,
"step": 1309
},
{
"epoch": 0.16307716265670014,
"eval_validation_agenda_digitale_loss": 1.6082433462142944,
"eval_validation_agenda_digitale_runtime": 173.0409,
"eval_validation_agenda_digitale_samples_per_second": 1.294,
"eval_validation_agenda_digitale_steps_per_second": 1.294,
"step": 1309
},
{
"epoch": 0.16307716265670014,
"eval_validation_legal_articles_loss": 1.4851994514465332,
"eval_validation_legal_articles_runtime": 277.1787,
"eval_validation_legal_articles_samples_per_second": 1.295,
"eval_validation_legal_articles_steps_per_second": 1.295,
"step": 1309
},
{
"epoch": 0.16307716265670014,
"eval_validation_leggepertutti_loss": 1.2786071300506592,
"eval_validation_leggepertutti_runtime": 30.9228,
"eval_validation_leggepertutti_samples_per_second": 1.294,
"eval_validation_leggepertutti_steps_per_second": 1.294,
"step": 1309
},
{
"epoch": 0.1863739001790859,
"grad_norm": 1.21875,
"learning_rate": 1e-05,
"loss": 1.2056,
"step": 1496
},
{
"epoch": 0.1863739001790859,
"eval_validation_loss": 1.031648874282837,
"eval_validation_runtime": 191.6944,
"eval_validation_samples_per_second": 1.294,
"eval_validation_steps_per_second": 1.294,
"step": 1496
},
{
"epoch": 0.1863739001790859,
"eval_validation_privacy_sources_loss": 1.031648874282837,
"eval_validation_privacy_sources_runtime": 191.7119,
"eval_validation_privacy_sources_samples_per_second": 1.294,
"eval_validation_privacy_sources_steps_per_second": 1.294,
"step": 1496
},
{
"epoch": 0.1863739001790859,
"eval_validation_agenda_digitale_loss": 1.6070060729980469,
"eval_validation_agenda_digitale_runtime": 173.203,
"eval_validation_agenda_digitale_samples_per_second": 1.293,
"eval_validation_agenda_digitale_steps_per_second": 1.293,
"step": 1496
},
{
"epoch": 0.1863739001790859,
"eval_validation_legal_articles_loss": 1.4841357469558716,
"eval_validation_legal_articles_runtime": 277.708,
"eval_validation_legal_articles_samples_per_second": 1.293,
"eval_validation_legal_articles_steps_per_second": 1.293,
"step": 1496
},
{
"epoch": 0.1863739001790859,
"eval_validation_leggepertutti_loss": 1.2782691717147827,
"eval_validation_leggepertutti_runtime": 30.9424,
"eval_validation_leggepertutti_samples_per_second": 1.293,
"eval_validation_leggepertutti_steps_per_second": 1.293,
"step": 1496
},
{
"epoch": 0.20967063770147162,
"grad_norm": 1.15625,
"learning_rate": 1e-05,
"loss": 1.1984,
"step": 1683
},
{
"epoch": 0.20967063770147162,
"eval_validation_loss": 1.027178406715393,
"eval_validation_runtime": 191.4569,
"eval_validation_samples_per_second": 1.295,
"eval_validation_steps_per_second": 1.295,
"step": 1683
},
{
"epoch": 0.20967063770147162,
"eval_validation_privacy_sources_loss": 1.027178406715393,
"eval_validation_privacy_sources_runtime": 191.6775,
"eval_validation_privacy_sources_samples_per_second": 1.294,
"eval_validation_privacy_sources_steps_per_second": 1.294,
"step": 1683
},
{
"epoch": 0.20967063770147162,
"eval_validation_agenda_digitale_loss": 1.6063356399536133,
"eval_validation_agenda_digitale_runtime": 173.2069,
"eval_validation_agenda_digitale_samples_per_second": 1.293,
"eval_validation_agenda_digitale_steps_per_second": 1.293,
"step": 1683
},
{
"epoch": 0.20967063770147162,
"eval_validation_legal_articles_loss": 1.4831825494766235,
"eval_validation_legal_articles_runtime": 277.5296,
"eval_validation_legal_articles_samples_per_second": 1.294,
"eval_validation_legal_articles_steps_per_second": 1.294,
"step": 1683
},
{
"epoch": 0.20967063770147162,
"eval_validation_leggepertutti_loss": 1.2776217460632324,
"eval_validation_leggepertutti_runtime": 30.9318,
"eval_validation_leggepertutti_samples_per_second": 1.293,
"eval_validation_leggepertutti_steps_per_second": 1.293,
"step": 1683
},
{
"epoch": 0.23296737522385735,
"grad_norm": 1.1484375,
"learning_rate": 1e-05,
"loss": 1.203,
"step": 1870
},
{
"epoch": 0.23296737522385735,
"eval_validation_loss": 1.0220788717269897,
"eval_validation_runtime": 191.5429,
"eval_validation_samples_per_second": 1.295,
"eval_validation_steps_per_second": 1.295,
"step": 1870
},
{
"epoch": 0.23296737522385735,
"eval_validation_privacy_sources_loss": 1.0220788717269897,
"eval_validation_privacy_sources_runtime": 191.6167,
"eval_validation_privacy_sources_samples_per_second": 1.294,
"eval_validation_privacy_sources_steps_per_second": 1.294,
"step": 1870
},
{
"epoch": 0.23296737522385735,
"eval_validation_agenda_digitale_loss": 1.605438470840454,
"eval_validation_agenda_digitale_runtime": 173.1034,
"eval_validation_agenda_digitale_samples_per_second": 1.294,
"eval_validation_agenda_digitale_steps_per_second": 1.294,
"step": 1870
},
{
"epoch": 0.23296737522385735,
"eval_validation_legal_articles_loss": 1.4823062419891357,
"eval_validation_legal_articles_runtime": 277.4211,
"eval_validation_legal_articles_samples_per_second": 1.294,
"eval_validation_legal_articles_steps_per_second": 1.294,
"step": 1870
},
{
"epoch": 0.23296737522385735,
"eval_validation_leggepertutti_loss": 1.2765312194824219,
"eval_validation_leggepertutti_runtime": 30.9074,
"eval_validation_leggepertutti_samples_per_second": 1.294,
"eval_validation_leggepertutti_steps_per_second": 1.294,
"step": 1870
},
{
"epoch": 0.2562641127462431,
"grad_norm": 1.296875,
"learning_rate": 1e-05,
"loss": 1.1827,
"step": 2057
},
{
"epoch": 0.2562641127462431,
"eval_validation_loss": 1.0178149938583374,
"eval_validation_runtime": 191.6897,
"eval_validation_samples_per_second": 1.294,
"eval_validation_steps_per_second": 1.294,
"step": 2057
},
{
"epoch": 0.2562641127462431,
"eval_validation_privacy_sources_loss": 1.0178149938583374,
"eval_validation_privacy_sources_runtime": 191.6258,
"eval_validation_privacy_sources_samples_per_second": 1.294,
"eval_validation_privacy_sources_steps_per_second": 1.294,
"step": 2057
},
{
"epoch": 0.2562641127462431,
"eval_validation_agenda_digitale_loss": 1.604758858680725,
"eval_validation_agenda_digitale_runtime": 173.0903,
"eval_validation_agenda_digitale_samples_per_second": 1.294,
"eval_validation_agenda_digitale_steps_per_second": 1.294,
"step": 2057
},
{
"epoch": 0.2562641127462431,
"eval_validation_legal_articles_loss": 1.4814982414245605,
"eval_validation_legal_articles_runtime": 277.3842,
"eval_validation_legal_articles_samples_per_second": 1.294,
"eval_validation_legal_articles_steps_per_second": 1.294,
"step": 2057
},
{
"epoch": 0.2562641127462431,
"eval_validation_leggepertutti_loss": 1.2762386798858643,
"eval_validation_leggepertutti_runtime": 30.9122,
"eval_validation_leggepertutti_samples_per_second": 1.294,
"eval_validation_leggepertutti_steps_per_second": 1.294,
"step": 2057
},
{
"epoch": 0.2795608502686288,
"grad_norm": 1.3046875,
"learning_rate": 1e-05,
"loss": 1.1921,
"step": 2244
},
{
"epoch": 0.2795608502686288,
"eval_validation_loss": 1.0138099193572998,
"eval_validation_runtime": 191.5601,
"eval_validation_samples_per_second": 1.295,
"eval_validation_steps_per_second": 1.295,
"step": 2244
},
{
"epoch": 0.2795608502686288,
"eval_validation_privacy_sources_loss": 1.0138099193572998,
"eval_validation_privacy_sources_runtime": 191.5465,
"eval_validation_privacy_sources_samples_per_second": 1.295,
"eval_validation_privacy_sources_steps_per_second": 1.295,
"step": 2244
},
{
"epoch": 0.2795608502686288,
"eval_validation_agenda_digitale_loss": 1.6038579940795898,
"eval_validation_agenda_digitale_runtime": 173.0095,
"eval_validation_agenda_digitale_samples_per_second": 1.295,
"eval_validation_agenda_digitale_steps_per_second": 1.295,
"step": 2244
},
{
"epoch": 0.2795608502686288,
"eval_validation_legal_articles_loss": 1.480303168296814,
"eval_validation_legal_articles_runtime": 277.4669,
"eval_validation_legal_articles_samples_per_second": 1.294,
"eval_validation_legal_articles_steps_per_second": 1.294,
"step": 2244
},
{
"epoch": 0.2795608502686288,
"eval_validation_leggepertutti_loss": 1.275497317314148,
"eval_validation_leggepertutti_runtime": 30.8976,
"eval_validation_leggepertutti_samples_per_second": 1.295,
"eval_validation_leggepertutti_steps_per_second": 1.295,
"step": 2244
},
{
"epoch": 0.30285758779101457,
"grad_norm": 1.2265625,
"learning_rate": 1e-05,
"loss": 1.1871,
"step": 2431
},
{
"epoch": 0.30285758779101457,
"eval_validation_loss": 1.009752631187439,
"eval_validation_runtime": 191.3996,
"eval_validation_samples_per_second": 1.296,
"eval_validation_steps_per_second": 1.296,
"step": 2431
},
{
"epoch": 0.30285758779101457,
"eval_validation_privacy_sources_loss": 1.009752631187439,
"eval_validation_privacy_sources_runtime": 191.5515,
"eval_validation_privacy_sources_samples_per_second": 1.295,
"eval_validation_privacy_sources_steps_per_second": 1.295,
"step": 2431
},
{
"epoch": 0.30285758779101457,
"eval_validation_agenda_digitale_loss": 1.602821946144104,
"eval_validation_agenda_digitale_runtime": 173.037,
"eval_validation_agenda_digitale_samples_per_second": 1.295,
"eval_validation_agenda_digitale_steps_per_second": 1.295,
"step": 2431
},
{
"epoch": 0.30285758779101457,
"eval_validation_legal_articles_loss": 1.4794570207595825,
"eval_validation_legal_articles_runtime": 277.2204,
"eval_validation_legal_articles_samples_per_second": 1.295,
"eval_validation_legal_articles_steps_per_second": 1.295,
"step": 2431
},
{
"epoch": 0.30285758779101457,
"eval_validation_leggepertutti_loss": 1.2749414443969727,
"eval_validation_leggepertutti_runtime": 30.8168,
"eval_validation_leggepertutti_samples_per_second": 1.298,
"eval_validation_leggepertutti_steps_per_second": 1.298,
"step": 2431
},
{
"epoch": 0.3261543253134003,
"grad_norm": 1.203125,
"learning_rate": 1e-05,
"loss": 1.1846,
"step": 2618
},
{
"epoch": 0.3261543253134003,
"eval_validation_loss": 1.0070545673370361,
"eval_validation_runtime": 191.5646,
"eval_validation_samples_per_second": 1.295,
"eval_validation_steps_per_second": 1.295,
"step": 2618
},
{
"epoch": 0.3261543253134003,
"eval_validation_privacy_sources_loss": 1.0070545673370361,
"eval_validation_privacy_sources_runtime": 191.5322,
"eval_validation_privacy_sources_samples_per_second": 1.295,
"eval_validation_privacy_sources_steps_per_second": 1.295,
"step": 2618
},
{
"epoch": 0.3261543253134003,
"eval_validation_agenda_digitale_loss": 1.6021575927734375,
"eval_validation_agenda_digitale_runtime": 172.9528,
"eval_validation_agenda_digitale_samples_per_second": 1.295,
"eval_validation_agenda_digitale_steps_per_second": 1.295,
"step": 2618
},
{
"epoch": 0.3261543253134003,
"eval_validation_legal_articles_loss": 1.4785128831863403,
"eval_validation_legal_articles_runtime": 277.2001,
"eval_validation_legal_articles_samples_per_second": 1.295,
"eval_validation_legal_articles_steps_per_second": 1.295,
"step": 2618
},
{
"epoch": 0.3261543253134003,
"eval_validation_leggepertutti_loss": 1.2738966941833496,
"eval_validation_leggepertutti_runtime": 30.8867,
"eval_validation_leggepertutti_samples_per_second": 1.295,
"eval_validation_leggepertutti_steps_per_second": 1.295,
"step": 2618
},
{
"epoch": 0.34945106283578603,
"grad_norm": 1.203125,
"learning_rate": 1e-05,
"loss": 1.1727,
"step": 2805
},
{
"epoch": 0.34945106283578603,
"eval_validation_loss": 1.0037845373153687,
"eval_validation_runtime": 191.6316,
"eval_validation_samples_per_second": 1.294,
"eval_validation_steps_per_second": 1.294,
"step": 2805
},
{
"epoch": 0.34945106283578603,
"eval_validation_privacy_sources_loss": 1.0037845373153687,
"eval_validation_privacy_sources_runtime": 191.5406,
"eval_validation_privacy_sources_samples_per_second": 1.295,
"eval_validation_privacy_sources_steps_per_second": 1.295,
"step": 2805
},
{
"epoch": 0.34945106283578603,
"eval_validation_agenda_digitale_loss": 1.6019808053970337,
"eval_validation_agenda_digitale_runtime": 172.902,
"eval_validation_agenda_digitale_samples_per_second": 1.296,
"eval_validation_agenda_digitale_steps_per_second": 1.296,
"step": 2805
},
{
"epoch": 0.34945106283578603,
"eval_validation_legal_articles_loss": 1.478790521621704,
"eval_validation_legal_articles_runtime": 277.2341,
"eval_validation_legal_articles_samples_per_second": 1.295,
"eval_validation_legal_articles_steps_per_second": 1.295,
"step": 2805
},
{
"epoch": 0.34945106283578603,
"eval_validation_leggepertutti_loss": 1.274503469467163,
"eval_validation_leggepertutti_runtime": 30.8879,
"eval_validation_leggepertutti_samples_per_second": 1.295,
"eval_validation_leggepertutti_steps_per_second": 1.295,
"step": 2805
},
{
"epoch": 0.3727478003581718,
"grad_norm": 1.25,
"learning_rate": 1e-05,
"loss": 1.1638,
"step": 2992
},
{
"epoch": 0.3727478003581718,
"eval_validation_loss": 1.001368522644043,
"eval_validation_runtime": 191.6637,
"eval_validation_samples_per_second": 1.294,
"eval_validation_steps_per_second": 1.294,
"step": 2992
},
{
"epoch": 0.3727478003581718,
"eval_validation_privacy_sources_loss": 1.001368522644043,
"eval_validation_privacy_sources_runtime": 191.6178,
"eval_validation_privacy_sources_samples_per_second": 1.294,
"eval_validation_privacy_sources_steps_per_second": 1.294,
"step": 2992
},
{
"epoch": 0.3727478003581718,
"eval_validation_agenda_digitale_loss": 1.6008728742599487,
"eval_validation_agenda_digitale_runtime": 173.1119,
"eval_validation_agenda_digitale_samples_per_second": 1.294,
"eval_validation_agenda_digitale_steps_per_second": 1.294,
"step": 2992
},
{
"epoch": 0.3727478003581718,
"eval_validation_legal_articles_loss": 1.4775311946868896,
"eval_validation_legal_articles_runtime": 277.4524,
"eval_validation_legal_articles_samples_per_second": 1.294,
"eval_validation_legal_articles_steps_per_second": 1.294,
"step": 2992
},
{
"epoch": 0.3727478003581718,
"eval_validation_leggepertutti_loss": 1.2725958824157715,
"eval_validation_leggepertutti_runtime": 30.9084,
"eval_validation_leggepertutti_samples_per_second": 1.294,
"eval_validation_leggepertutti_steps_per_second": 1.294,
"step": 2992
},
{
"epoch": 0.3960445378805575,
"grad_norm": 1.2109375,
"learning_rate": 1e-05,
"loss": 1.1752,
"step": 3179
},
{
"epoch": 0.3960445378805575,
"eval_validation_loss": 0.9985631704330444,
"eval_validation_runtime": 191.601,
"eval_validation_samples_per_second": 1.294,
"eval_validation_steps_per_second": 1.294,
"step": 3179
},
{
"epoch": 0.3960445378805575,
"eval_validation_privacy_sources_loss": 0.9985631704330444,
"eval_validation_privacy_sources_runtime": 191.6183,
"eval_validation_privacy_sources_samples_per_second": 1.294,
"eval_validation_privacy_sources_steps_per_second": 1.294,
"step": 3179
},
{
"epoch": 0.3960445378805575,
"eval_validation_agenda_digitale_loss": 1.6005090475082397,
"eval_validation_agenda_digitale_runtime": 173.0062,
"eval_validation_agenda_digitale_samples_per_second": 1.295,
"eval_validation_agenda_digitale_steps_per_second": 1.295,
"step": 3179
},
{
"epoch": 0.3960445378805575,
"eval_validation_legal_articles_loss": 1.4775700569152832,
"eval_validation_legal_articles_runtime": 277.3007,
"eval_validation_legal_articles_samples_per_second": 1.295,
"eval_validation_legal_articles_steps_per_second": 1.295,
"step": 3179
},
{
"epoch": 0.3960445378805575,
"eval_validation_leggepertutti_loss": 1.2739975452423096,
"eval_validation_leggepertutti_runtime": 30.91,
"eval_validation_leggepertutti_samples_per_second": 1.294,
"eval_validation_leggepertutti_steps_per_second": 1.294,
"step": 3179
},
{
"epoch": 0.41934127540294325,
"grad_norm": 1.1328125,
"learning_rate": 1e-05,
"loss": 1.1764,
"step": 3366
},
{
"epoch": 0.41934127540294325,
"eval_validation_loss": 0.997268557548523,
"eval_validation_runtime": 191.4685,
"eval_validation_samples_per_second": 1.295,
"eval_validation_steps_per_second": 1.295,
"step": 3366
},
{
"epoch": 0.41934127540294325,
"eval_validation_privacy_sources_loss": 0.997268557548523,
"eval_validation_privacy_sources_runtime": 191.462,
"eval_validation_privacy_sources_samples_per_second": 1.295,
"eval_validation_privacy_sources_steps_per_second": 1.295,
"step": 3366
},
{
"epoch": 0.41934127540294325,
"eval_validation_agenda_digitale_loss": 1.600382924079895,
"eval_validation_agenda_digitale_runtime": 172.9802,
"eval_validation_agenda_digitale_samples_per_second": 1.295,
"eval_validation_agenda_digitale_steps_per_second": 1.295,
"step": 3366
},
{
"epoch": 0.41934127540294325,
"eval_validation_legal_articles_loss": 1.4765703678131104,
"eval_validation_legal_articles_runtime": 277.255,
"eval_validation_legal_articles_samples_per_second": 1.295,
"eval_validation_legal_articles_steps_per_second": 1.295,
"step": 3366
},
{
"epoch": 0.41934127540294325,
"eval_validation_leggepertutti_loss": 1.272763967514038,
"eval_validation_leggepertutti_runtime": 30.8904,
"eval_validation_leggepertutti_samples_per_second": 1.295,
"eval_validation_leggepertutti_steps_per_second": 1.295,
"step": 3366
}
],
"logging_steps": 187,
"max_steps": 8026,
"num_input_tokens_seen": 0,
"num_train_epochs": 1,
"save_steps": 187,
"stateful_callbacks": {
"TrainerControl": {
"args": {
"should_epoch_stop": false,
"should_evaluate": false,
"should_log": false,
"should_save": true,
"should_training_stop": false
},
"attributes": {}
}
},
"total_flos": 1.987621253289817e+19,
"train_batch_size": 1,
"trial_name": null,
"trial_params": null
}