lnp_ml/models/pretrain/llm/pretrain_history.json

166 lines
2.8 KiB
JSON

{
"train": [
{
"loss": 0.8493162816794767,
"n_samples": 8236
},
{
"loss": 0.719371718651696,
"n_samples": 8236
},
{
"loss": 0.6663210250440879,
"n_samples": 8236
},
{
"loss": 0.6313854984544677,
"n_samples": 8236
},
{
"loss": 0.6050180681936367,
"n_samples": 8236
},
{
"loss": 0.5724720832732999,
"n_samples": 8236
},
{
"loss": 0.5582639975939172,
"n_samples": 8236
},
{
"loss": 0.5438417370832331,
"n_samples": 8236
},
{
"loss": 0.5246556448728035,
"n_samples": 8236
},
{
"loss": 0.5136562741283307,
"n_samples": 8236
},
{
"loss": 0.49826916715020747,
"n_samples": 8236
},
{
"loss": 0.4817734892186521,
"n_samples": 8236
},
{
"loss": 0.4816470630629772,
"n_samples": 8236
},
{
"loss": 0.4610506330895852,
"n_samples": 8236
},
{
"loss": 0.4502680387063881,
"n_samples": 8236
},
{
"loss": 0.4463582375259177,
"n_samples": 8236
},
{
"loss": 0.40212532723772115,
"n_samples": 8236
},
{
"loss": 0.39290236833316716,
"n_samples": 8236
},
{
"loss": 0.3818181393041143,
"n_samples": 8236
},
{
"loss": 0.38346117517796463,
"n_samples": 8236
}
],
"val": [
{
"loss": 0.7823408520697236,
"n_samples": 1454
},
{
"loss": 0.7572638167803685,
"n_samples": 1454
},
{
"loss": 1.3016815285571175,
"n_samples": 1454
},
{
"loss": 0.7050852468941857,
"n_samples": 1454
},
{
"loss": 0.7743825887879446,
"n_samples": 1454
},
{
"loss": 2.98265173149896,
"n_samples": 1454
},
{
"loss": 0.7438884984184329,
"n_samples": 1454
},
{
"loss": 2.6711699774045563,
"n_samples": 1454
},
{
"loss": 0.7234079081057846,
"n_samples": 1454
},
{
"loss": 0.6671479742005569,
"n_samples": 1454
},
{
"loss": 1.0528025386913935,
"n_samples": 1454
},
{
"loss": 0.7759058258228486,
"n_samples": 1454
},
{
"loss": 1.0733910772790607,
"n_samples": 1454
},
{
"loss": 0.7092724900790061,
"n_samples": 1454
},
{
"loss": 0.8837297995775897,
"n_samples": 1454
},
{
"loss": 0.874810852243615,
"n_samples": 1454
},
{
"loss": 0.7093387734775202,
"n_samples": 1454
},
{
"loss": 0.7450522419675016,
"n_samples": 1454
},
{
"loss": 0.7102828234065186,
"n_samples": 1454
},
{
"loss": 0.702132918893717,
"n_samples": 1454
}
]
}