slowrun-checkpoints / progress.json
ugonfor's picture
Upload progress.json with huggingface_hub
d9b4e34 verified
Raw
History Blame Contribute Delete
3.5 kB
{
"models": [
{
"idx": 0,
"seed": 42,
"hparams": {
"matrix_lr": 0.08,
"scalar_lr": 0.5,
"weight_decay": 1.6,
"dropout": 0.1,
"num_epochs": 16,
"ema_decay": 0.999
},
"val_bpb": 1.095866685089716,
"val_loss": 3.3719550461065575,
"paths": [
"/home/jovyan/slowrun/checkpoints/model_0.pt"
]
},
{
"idx": 1,
"seed": 43,
"hparams": {
"matrix_lr": 0.08,
"scalar_lr": 0.5,
"weight_decay": 1.8,
"dropout": 0.1,
"num_epochs": 16,
"ema_decay": 0.999
},
"val_bpb": 1.0963785699410264,
"val_loss": 3.373510341956967,
"paths": [
"/home/jovyan/slowrun/checkpoints/model_1.pt"
]
},
{
"idx": 2,
"seed": 44,
"hparams": {
"matrix_lr": 0.08,
"scalar_lr": 0.5,
"weight_decay": 1.6,
"dropout": 0.12,
"num_epochs": 16,
"ema_decay": 0.999
},
"val_bpb": 1.0952586997499123,
"val_loss": 3.3700663582223362,
"paths": [
"/home/jovyan/slowrun/checkpoints/model_2.pt"
]
},
{
"idx": 3,
"seed": 45,
"hparams": {
"matrix_lr": 0.08,
"scalar_lr": 0.5,
"weight_decay": 1.4,
"dropout": 0.1,
"num_epochs": 16,
"ema_decay": 0.999
},
"val_bpb": 1.095667712804597,
"val_loss": 3.3713214811731556,
"paths": [
"/home/jovyan/slowrun/checkpoints/model_3.pt"
]
},
{
"idx": 4,
"seed": 46,
"hparams": {
"matrix_lr": 0.08,
"scalar_lr": 0.5,
"weight_decay": 1.8,
"dropout": 0.08,
"num_epochs": 16,
"ema_decay": 0.999
},
"val_bpb": 1.096950224608351,
"val_loss": 3.3753217853483606,
"paths": [
"/home/jovyan/slowrun/checkpoints/model_4.pt"
]
},
{
"idx": 5,
"seed": 47,
"hparams": {
"matrix_lr": 0.07,
"scalar_lr": 0.5,
"weight_decay": 1.6,
"dropout": 0.1,
"num_epochs": 16,
"ema_decay": 0.999
},
"val_bpb": 1.0949341978294176,
"val_loss": 3.3690793897284834,
"paths": [
"/home/jovyan/slowrun/checkpoints/model_5.pt"
]
},
{
"idx": 6,
"seed": 48,
"hparams": {
"matrix_lr": 0.08,
"scalar_lr": 0.5,
"weight_decay": 2.0,
"dropout": 0.12,
"num_epochs": 16,
"ema_decay": 0.999
},
"val_bpb": 1.0980855906776907,
"val_loss": 3.3788153976690576,
"paths": [
"/home/jovyan/slowrun/checkpoints/model_6.pt"
]
},
{
"idx": 7,
"seed": 49,
"hparams": {
"matrix_lr": 0.09,
"scalar_lr": 0.5,
"weight_decay": 1.6,
"dropout": 0.1,
"num_epochs": 16,
"ema_decay": 0.999
},
"val_bpb": 1.096767,
"val_loss": 3.374763,
"paths": [
"/home/jovyan/slowrun/checkpoints/model_7.pt"
]
}
],
"ensembles": [
{
"num_checkpoints": 2,
"ensemble_bpb": 1.0741511153509162,
"ensemble_loss": 3.305124651233176
},
{
"num_checkpoints": 4,
"ensemble_bpb": 1.0611368083045234,
"ensemble_loss": 3.2650819475755295
},
{
"num_checkpoints": 6,
"ensemble_bpb": 1.0571252225423302,
"ensemble_loss": 3.2527488215884826
}
]
}