16 lines
262 B
JSON
16 lines
262 B
JSON
{
|
|
"train_batch_size": 2,
|
|
"gradient_accumulation_steps": 1,
|
|
"optimizer": {
|
|
"type": "Adam",
|
|
"params": {
|
|
"lr": 0.00015
|
|
}
|
|
},
|
|
"gradient_clipping": 1.0,
|
|
"fp16": {
|
|
"enabled": true,
|
|
"loss_scale": 0
|
|
}
|
|
}
|